Aliment Pharmacol Ther 2011; 33: 1133–1142
Background Crohn’s disease incidence rates have stabilised in industrialised countries since the 1980s. Conversely, a continuing increase in childhood-onset Crohn’s disease incidence has been reported.
Aim To confirm trends in inflammatory bowel disease (IBD) incidence in northern France over an extended time period (1988–2007) with a focus on childhood-onset Crohn’s disease.
Methods The IBD patients recorded in the EPIMAD registry between 1988 and 2007 were included. Standardised incidence rates were calculated for Crohn’s disease and ulcerative colitis in the entire population, and separately according to age. Evolution of phenotypes at diagnosis was also studied.
Results A total of 12 084 incident IBD cases (7428 Crohn’s disease and 4656 ulcerative colitis) were recorded. Crohn’s disease incidence rates increased from 5.2 cases/100 000 persons in 1988–1990 to 6.7 in 2006–2007 (+29%), stabilising after a peak at 7.1 in 1997–1999. Crohn’s disease incidence rates in the 10–19-year age category increased by 71%, from 6.5 (1988–1990) to 11.1 (2006–2007). The frequency of initial ileo-colonic localisation increased from 52.9% in 1988–1990 to 68.6% in 2006–2007 (P < 0.0001). Ulcerative colitis incidence rates decreased during the same period.
Conclusions From 1988 to 2007, Crohn’s disease incidence increased by 29% in northern France and by 71% in the 10–19-year-old age group. Consequently, studies on Crohn’s disease risk factors should focus on the population under 20 years of age.
Inflammatory bowel diseases (IBD) represent a major burden for developed countries. It has been estimated that as many as 1.4 million persons may be afflicted with IBD in the US and Canada, and 2.2 million in Europe1 Crohn’s disease (CD) and ulcerative colitis (UC) are the major IBD that challenge gastroenterologists. Although significant advances have been made towards a better understanding of the pathophysiology of IBD,2, 3 their aetiology remains unclear.1 Epidemiological studies of geographic and temporal variations in IBD provide important information on the natural history, health care burden and causal mechanisms of the disease. In particular, population-based epidemiological studies are more likely to reflect the true spectrum of illness than studies performed on patients seen at referral centres. Indeed, epidemiological studies have shown a continuing increase in4–6 or stabilisation7–9 of CD incidence rates in developed countries over the past five decades. In contrast, several studies have described a continuing increase in incidence rates of CD in children and adolescents,10–16 suggesting that ongoing risk factors may exist in these populations. Nevertheless, these changes in the incidence of juvenile-onset CD remain subject to debate and need to be confirmed in larger population-based registries. The EPIMAD study records all incident cases of IBD since 1988 in northern France. Previously published studies covering the 1988–1999 period reported a significant increase in incidences rates of CD in the overall population,17 as well as a nonsignificant increase in the paediatric population.18 The aim of the present study was therefore to conduct an updated analysis of the evolution of incidence rates and digestive sites of CD and UC for the 1988–2007 periods in a large population registry which included infants and young adults.
Patients and methods
Cases included all patients from the EPIMAD registry who had a diagnosis of definite or probable IBD, including cases of CD and UC, between January 1988 and December 2007. The study area was the northern part of France, which has 5 790 526 inhabitants (1999 national population census) residing in four regions: Nord, Pas-de-Calais, Somme and Seine-Maritime. This part of France represents 9.3% of the total French population and is composed of four sectors: (i) Nord, with 2 554 449 inhabitants and a population density of 445/km2; (ii) Pas-de-Calais, with 1 441 422 inhabitants and a population density of 216/km2; (iii) Somme, with 555 479 inhabitants and a population density of 90/km2; and (iv) Seine-Maritime, with 1 239 176 inhabitants and a population density of 197/km2. Both rural and urban populations can be found in these areas (ratio urban/rural = 8.9 in the Nord, 4.5 in the Pas-de-Calais, 1.4 in the Somme, and 3.0 in the Seine-Maritime). This region is a well-defined geographic entity bordering Belgium and the North Sea (Figure 1). The population is stable – i.e. the percentages of the population moving per year for each area were 0.8 for Nord and Pas-de-Calais, 1.1 for Somme and 0.9 for Seine-Maritime (French National Statistical Institute).
The methodology of the EPIMAD registry has been previously described in detail.17–19 Interviewing practitioners collected data on all patients diagnosed between 1988 and 2007 from all adult and paediatric gastroenterologists (GE) (total: 248) practising in the private or public sector. Only patients who had been residents of the study areas at the time of diagnosis of IBD were included. Each GE reported any patient consulting for the first time with clinical symptoms compatible with IBD, and was contacted by phone at least three times a year by an interviewing practitioner. This interviewer went to the GE’s consulting room and collected data from charts in a standardised questionnaire for each new case. Main data collected included age, gender, year of diagnosis, interval between onset of symptoms and diagnosis, and clinical, radiological, endoscopic and histological findings at the time of diagnosis. A final diagnosis of CD or UC was made by two expert gastroenterologists and was recorded as definite, probable or possible according to previously published criteria.19 For the purpose of this study, only patients with definite or probable IBD were considered.
For the study of phenotypes of CD, the site of disease was reported only for patients who underwent a complete bowel investigation (images of small and large bowel). The localisations of the lesions were coded according to the Montreal classification20 with modification for CD: L1 referred to ileal localisation of lesions, L2 to colonic localisation and L3 to ileo-colonic localisation, with the addition that an ileal localisation with cecal involvement was considered as an L3 localisation. L4 referred to upper gastrointestinal location of the lesions. For UC localisation, ulcerative proctitis (E1) was defined as involvement limited to the rectum (i.e. the proximal extent of inflammation was distal to the rectosigmoid junction), left-sided UC (E2) was defined as involvement limited to the portion of the colorectum distal to the splenic flexure, and extensive UC (E3) was defined as involvement extending proximal to the splenic flexure. Extra intestinal manifestations (EIMs) included joints, skin, ocular and hepato-biliary manifestations.
Incidence rates were computed as the ratio of the number of incident cases with respect to the number of person-years for both sexes. Yearly estimations of population based on a procedure mixing exhaustive census before 2004 and random sampling after 2004 were obtained from the French Statistics Institute (INSEE). Three-year periods were used to ensure better statistical precision and robustness. Age standardisation was performed using European standard population weights for 5-year age groups. Incidence rates were studied in the overall population and according to 20-year age brackets. Finally, 10-year category incidence rates were analysed for patients diagnosed prior to the age of 20 years. Exact confidence intervals for standardised incidence rates were approximated using a gamma distribution.21 Contributions of age, sex, time period and region were tested for CD in a generalised linear model using Poisson distribution. Second-order interactions between these variables were selected using Akahike Information Criteria (AIC).22 Linearity of the evolution of incidence rates of CD was assessed with a likelihood ratio test using nested models. Changes in CD localisation and time were assessed with appropriate Chi-squared tests. For descriptions of age at diagnosis (years) and time intervals between symptoms and diagnosis (months), median and interquartile ranges (25th and 75th percentiles) are given. Comparisons of distribution of these variables across time were performed using the Wilcoxon, Kruskal–Wallis or Chi-squared tests for trend. Analyses were performed with R 220.127.116.11
During the 1988–2007 period, 12 052 patients satisfying the criteria defined above were recorded in the EPIMAD registry, including 61% (n = 7409) of CD cases and 39% (n = 4643) of UC cases. Median age at diagnosis was 30 years (interquartile range IQR: 22-42) and was significantly higher in patients with UC than in patients with CD (35 vs. 26 years, respectively; P < 0.0001). Median delay between first symptoms and diagnosis was 3 months (IQR: 1-6) and was significantly higher in CD cases than in UC cases (3 vs. 2 months, respectively; P < 0.0001). Five hundred and sixty cases of indeterminate colitis (IC) were recorded.
Incidence of CD
During the 1988–2007 period, 7409 patients were diagnosed with either definite (35%n = 2570) or probable (65%n = 4839) CD, with a predominance of women (56%n = 4180). Median age at diagnosis was 26 years (IQR: 20-38) and was not statistically different between men and women (P = 0.8). Median interval between first symptoms and diagnosis was 3 months (IQR: 1–8) for the entire period, with a significant decrease in the proportion of intervals longer than 9 months, from 28% in 1988–1990 to 22% in 2006–2007 (P < 0.0001) (Table 1).
|Age at diagnosis (years)*||27 (20–38)||27 (20–38)||27 (21–38)||27 (20–38)||27 (20–38)||26 (20–39)||25 (19–37)|
|Interval before diagnosis (months)*||3 (1–12)||3 (1–11)||3 (1–8)||3 (1–7)||3 (1–6)||3 (1–6)||3 (1–8)|
|Family history of IBD % (n)†||9.4 (70)||11 (95)||11.6 (99)||14.4 (150)||13.9 (136)||15.6 (163)||14.8 (107)|
|EIMs % (n)†||15.6 (121)||14.5 (146)||11 (113)||13.7 (163)||11.4 (133)||9.9 (118)||11.8 (90)|
|Complete gut investigation % (n)†||76.9 (618)||72 (752)||73.3 (809)||79.4 (997)||79 (953)||81.2 (987)||86.5 (674)|
|Disease localisation % (n)†|
|L1||18.4 (100)||14.8 (102)||15.9 (117)||16 (144)||18 (144)||15 (118)||15.5 (84)|
|L2||22.6 (123)||16.6 (115)||12.6 (93)||13 (117)||9 (72)||9 (71)||5.4 (29)|
|L3||43.9 (239)||48.3 (334)||49.3 (363)||46.1 (416)||44.4 (356)||48.3 (380)||45 (244)|
|L1 + L4||3.7 (20)||3.8 (26)||5.3 (39)||4.7 (42)||3.6 (29)||5 (39)||7.6 (41)|
|L2 + L4||3.3 (18)||3 (21)||2.7 (20)||3.2 (29)||2.9 (23)||1.7 (13)||3.3 (18)|
|L3 + L4||8.1 (44)||13.5 (93)||14.1 (104)||17.1 (154)||22.2 (178)||21 (165)||23.2 (126)|
|Anoperineal lesions % (n)†||17.8 (135)||15.4 (148)||12.1 (122)||11.2 (132)||11.4 (128)||12.2 (141)||12 (91)|
During the 1988–2007 period, incidence rate of CD was 6.4 [95% confidence interval (CI) 6.2–6.5]. CD incidence rate increased significantly, from 5.2 cases per 100 000 person-years in 1988–1990 to 6.7 in 2006–2007 (test for trend, P < 0.0001) (Figure 2). Rates reached a maximum of 7.1 in 1997–1999 and then remained stable (test for nonlinearity, P = 0.0003). CD incidence rate was significantly higher in women (7.1; 95% CI 6.9–7.3) than in men (5.7; 95% CI 5.5–5.9) (P < 0.0001), but analyses performed in each age group showed that this difference was present only in patients aged 20–24 years (P for interaction between age and sex <0.0001) (Figure 3). This difference between women and men remained stable during the study period (P for interaction = 0.21). Evolutions in incidence rates did not differ according to region (interaction, P = 0.08).
Evolution of incidence differed according to age group (Figure 4a). Incidence rates in the 0–19 age group increased steadily over the entire study period, from 3.4 cases per 100 000 person-years in 1988–1990 to 5.9 in 2006–2007 (P for trend <0.0001). Evolution of incidence rates in the 0–19 age group did not differ according to gender or region (interaction, P = 0.10 and P = 0.22, respectively). Further analyses showed that the 10–19-year-old age group was mainly responsible for this increase, since the incidence rates increased continuously, from 6.5 cases per 100 000 person-years in 1988–1990 to 11.1 in 2006–2007 (P for trend <0.0001), whereas the incidence rates in the 0–9-year-old age group only slightly increased during the same period (P for trend = 0.02, Figure 4b). Furthermore, the incidence rate in the age group 20–29 was similar to incidence rates in the overall population, with stabilisation beginning from the 1997–1999 period, whereas incidence rates in patients aged 30 and over remained stable or only slightly increased.
During the 1988–2007 period, 5790 patients underwent complete bowel investigation. The proportion of complete bowel investigations significantly increased, from 76.9% in 1988–1990 to 86.5% in 2006–2007 (P for trend <0.0001). An ileo-colonic localisation (L3 and L3 + L4) was predominant in the overall population and increased during the study period, from 52.0% in 1988–1990 to 68.3% in 2006–2007 (P for trend = 0.0004). Extended phenotypes (L3 + L4) significantly increased, from 8.1% in 1988–1990 to 23.2% in 2006–2007 (P for trend <0.0001), whereas an ileo-colonic localisation without upper GI lesions (L3) remained stable. Colonic site forms (L2) significantly decreased, from 22.6% in 1988–1990 to 5.4% in 2006–2007 (P for trend <0.0001), whereas other sites remained stable (Table 1).
The proportion of patients with complete bowel investigation was stable in the 0–19 age group (test for trend, P = 0.12) and increased from 77.9% to 88.8% (test for trend, P < 0.0001), from 68.6% to 83.9% (test for trend, P = 0.003) and from 52.5% to 74.4% (test for trend, P = 0.013) in the 20–39, 40–59 and 60 and over age groups, respectively. Similar to the overall population, ileo-colonic (L3 and L3 + L4) forms of CD were predominant in children and adolescents, but a higher proportion of extended lesions (L3 + L4) than in other age groups was observed. The decreasing proportion of pure L2 localisations observed in the overall population was found in the 20–39-year-old and the 40–59-year-old age groups, as it decreased from 22.5% to 5.8% (test for trend, P < 0.0001) and from 40.3% to 4.8% (test for trend, P = 0.004), respectively, but not in the other age groups.
Incidence of UC
From 1988 to 2007, 4643 incident cases of UC were recorded in the EPIMAD study, with a predominance of men (55%; n = 2566). Median age at diagnosis was 35 (IQR: 26-47) years and was significantly higher in men than in women: 38 vs. 32, respectively (P < 0.0001). The median interval between first symptoms and diagnosis was 2 months (IQR: 1–6) for the entire period, with stable proportions over the time (Table 2). During the total period, UC incidence rate was 4.1 (95% CI 4.0–4.2) and was significantly higher in men (4.7; 95% CI 4.5–4.9) than in women (3.6; 95% CI 3.4–3.7) (P < 0.0001) (Figure 2). This difference in UC incidence rates according to gender was observed for patients over 35 (Figure 5). A significant decrease in UC incidence rate was observed from 4.3 (95% CI 4.0–4.7) in 1988–1990 to 3.4 (95% CI 3.1–3.7) in 2006–2007 (test for trend, P = 0.001). Time trends of UC incidence rates differed significantly according to age group (interaction, P < 0.0001) with a more important decrease in the patients aged 60 and over than the other age groups.
|Age at diagnosis (years)*||34 (26–48)||36 (27–49)||34 (26–47)||35 (26–47)||36 (27–48)||36 (27–48)||34 (23–47)|
|Interval before diagnosis (months)*||2 (1–6)||2 (1–6)||2 (1–6)||2 (1–4)||2 (1–4)||2 (1–5)||3 (2–6)|
|Family history of IBD % (n)†||3.9 (23)||3.9 (24)||6.1 (35)||7.1 (42)||7.3 (42)||7.8 (52)||7 (26)|
|EIMs % (n)†||4.9 (30)||3.3 (24)||2.8 (19)||2.9 (20)||2.3 (15)||2.8 (21)||3.1 (12)|
|Disease localisation % (n)†|
|E1||28.8 (120)||32.4 (164)||28.3 (141)||27.2 (150)||28 (144)||27.1 (163)||31.3 (101)|
|E2||45.7 (190)||42.9 (217)||43.6 (217)||45.1 (249)||46.6 (240)||47.4 (285)||39.3 (127)|
|E3||25.5 (106)||24.7 (125)||28.1 (140)||27.7 (153)||25.4 (131)||25.5 (153)||29.4 (95)|
During the 1988–2007 period, 3411 patients underwent complete colonic investigation. The proportion of complete colonic investigations carried out significantly increased, from 66.1% in 1988–1990 to 82.6% in 2006–2007 (P for trend <0.0001). E2 localisations constituted most UC cases, followed by E1 and E3 localisations (45%, 29% and 26%, respectively). These proportions remained stable during the study period (Table 2).
In Northern France, from 1988 to 2007, incidence rates of CD stabilised following an initial increase, while those of UC decreased during that period. Conversely, CD incidence rates increased in the 0–19-year-old age category, mainly due to the 10–19-year-old category whereas they remained stable or stabilised after initial increase in the other age groups. For CD localisation at diagnosis, ileo-colonic forms (L3) of CD were predominant in the overall population, with an increase of extended forms (at least three affected segments). In the 0–19-year-old category, a higher proportion of L3 + L4 localisations than in the whole population was observed, with similar evolution. Incidence rates of CD have stabilised in northern France since 1999. These results are coherent with previous studies in Minnesota,8 the United Kingdom9 and Sweden.7 Conversely, other studies in Denmark4, 5 and Greece6 have shown a continuing increase. Since only definite and probable cases of CD were included in the EPIMAD study, underestimation of incidence rates due to incomplete CD case ascertainment cannot be ruled out. Potential IBD cases likely to be reclassified as definite or probable CD, such as possible CD, UC or IC cases, represented 2436 patients in the entire database. Indeed, only continuous surveillance of these populations will enable us to more precisely determine the stabilisation of incidence rates of CD. It might be speculated that accuracy of case recording has improved with time in the EPIMAD registry following a learning curve from the time of registry creation. However, in our opinion, this is unlikely since, in this case, the same variation (i.e. modification in the slope beginning in 1997–1999) in incidence rates of UC would have been observed. Our earlier data17 as well as the present report did not support this hypothesis; indeed, a significant decrease in UC incidence over time was observed.
Our main result, indicating that incidence rates of CD continue to increase in young persons in Northern France, is of interest. Indeed, there existed stable incidence rates in the 0–9 year age group but increasing incidence rates in the 10–19-year-old age group. Similar results have been observed in Norway14 and Scotland,11 and in other studies in Finland,15 Sweden,10, 12 Australia13 and Canada.16 Those studies showed an overall increase in the incidence of childhood-onset CD. Better diagnostic procedures and greater awareness by parents of the risk of CD may have led to reducing the time lapse between symptoms and diagnosis, thereby increasing the incidence rates of CD in younger patients. In the EPIMAD study, the median time between symptoms and diagnosis was 3 months, which is low. Even if a decreasing proportion of intervals longer than 9 months had been observed over time, it would not easily confirm a decrease of several years of age at diagnosis. We therefore feel that a true increase in the incidence rates of CD is occurring in adolescents in Northern France. Moreover, if our observations had relied only on advances in diagnosis, then this would have impacted the incidence rate of the older age groups, which was not the case here. The increase in CD incidence in the 10–19 years group could also reflect an earlier presentation of CD in predisposed patients as proportion of familial history of CD increased and at the same time proportions of extra intestinal manifestations and anoperineal lesions decreased between 1988 and 2007. It might be hypothesised that, in Northern France over time, a change in smoking habits, a well-known risk factor for CD,24 would explain the differential evolution in incidence rates of CD according to age. However, accurate regional data supporting this hypothesis are scarce. At a national level, two population-based studies concerning teenagers showed a 30% decrease in the proportion of 17-year-old subjects reporting daily tobacco consumption, from 41.1% in 2000 to 28.9% in 2008,25 and a 45% decrease among 15- and 16-year-old subjects, from 31% in 1999 to 17% in 2007.26 Moreover, other national data on smoking habits indicate a decrease in the prevalence of smoking in every age group27, 28 during the study period. Change of diet has also been incriminated as a risk factor for IBD. A recent French cohort study of middle-aged women showed that protein intake was associated with a higher risk of IBD, CD in particular.29
Analysis of the localisations of CD lesions at diagnosis showed a predominance of ileo-colonic (L3) forms, with increasing proportions of extended forms, including ileo-colonic and upper gastrointestinal lesions (L3 + L4), from 1988 to 2007. These results are unusual compared to the literature. Some,4–7, 9 but not all,30 population-based studies reported a higher proportion of pure colonic localisations (L2) in CD patients. A possible explanation for these results might lie in the fact that we reported CD sites only in patients who had undergone a complete bowel investigation (with imaging of both large and small bowel). In other studies, the proportion of pure colonic localisations (L2) may have been overestimated; as they were calculated from the entire population regardless of whether or not small bowel investigation had been performed. Moreover, the recent use of wireless capsule endoscopy (WCE) may have impacted the proportion of patients diagnosed with extensive small bowel involvement. In France, however, use of WCE remained scarce before 2007 so it is unlikely to have influenced diagnosis of more extensive forms of CD.
Among patients aged 0 to 19, an ileo-colonic (L3) localisation constituted most CD lesions at diagnosis and this was stable over time. This is coherent with recent studies reporting extensive lesions of CD in childhood at diagnosis, either an ileo-colonic localisation14, 31, 32 or an association of ileo-colonic and upper gastrointestinal lesions.33 These differences in phenotypes among adults, children and adolescents may reflect the influence of distinct causal factors or a more systematic investigation of the gastrointestinal tract in younger patients. One strength of this study lies in the number of registered cases, which enabled precise estimates of incidence rates of CD and adjustment for covariates in statistical models. Systematic follow-up of newly diagnosed patients at 18 months allowed for potential reclassification and well-documented final diagnoses of CD. In contrast, questions concerning the completeness of case registration may be raised. The EPIMAD registry relied only on gastroenterologists’ declarations of incident cases, thereafter confirmed by experts. In France, gastroenterologists have a monopoly in their profession, especially in terms of administrative authorisation to perform endoscopy. As a consequence, there is a strong probability that each new IBD patient will consult a gastroenterologist and therefore be included in the study.
In Northern France, incidence rates of CD continue to rise among younger patients, with a more extensive phenotype than in adults. These findings suggest that different causal factors may exist in patients aged 10–20 years. It is therefore of great importance to focus aetiological research on comparisons of trends in incidence rates and phenotypes between adult- and childhood-onset CD.
The authors wish to thank the interviewing practitioners who collected data: B. Lemaire, N. Guillon, I. Rousseau, A Pétillon, B. Turck, P. Fosse, M. Leconte, C. Le Gallo, C. Dias and S. Richon. The authors thank all gastroenterologists and other specialists who participated in this study. Jean-Frédéric Colombel is the guarantor of the article. Declaration of personal interests: None. Declaration of funding interests: EPIMAD is organised under an agreement between the Institut National de la Santé et de la Recherche Médicale (INSERM) and the Institut National de Veille Sanitaire (InVS), and also received financial support from the François Aupetit Association, Lion’s Club of Northern France, Ferring Laboratories, Astra-Zeneca Company (IRMAD), the Société Nationale Française de Gastroentérologie and Lille University Hospital.