Changes in incidence and clinical features of inflammatory bowel disease in Cardiff, UK over 50 years: an update for 2005–2016

Abstract Introduction Population-based studies of inflammatory bowel disease (IBD) in Cardiff have recorded data back to 1930 for Crohn’s disease (CD) and 1968 for ulcerative colitis (UC). This study compares incidence and phenotype for 2005–2016 with past data. Methods All new IBD cases resident in the Cardiff at diagnosis were collected retrospectively for the 12-year period 2005–2016, and compared with previous Cardiff data for trends in incidence and phenotype. Overall incidence was age/sex corrected to the UK population. Results There were 991 new patients: 34% had CD, 5.4% IBD unclassified (IBD-U) and 60.5% had UC. The corrected incidence of CD was 7.7 per 100,000 person years [95% CI 6.9–8.6]. CD incidence is significantly higher than previous Cardiff studies, but the annual percentage change (APC) for 1980–2016 of 0.06; [95%CI −0.02 to 0.14] is not significant, with a previous higher APC for 1953–1980 of 0.18, [95%CI 0.13 to 0.23]. Uncorrected IBD-U incidence was 1.3 per 100,000 person years [95% CI 1.0–1.7]. UC corrected incidence was 14.4 per 100,000 person years [95% CI 13.3–15.6]. Incidence of UC is greater than in previous studies but did not increase during the current 12-year period. CD distribution at diagnosis continues to change as in previous Cardiff studies, with further increase in colonic disease and ileocolonic, (42% L2, 28% L3) and reduction in isolated terminal ileal disease (29% L1). Conclusions Incidence of both CD and UC are no longer rising significantly, but the location of CD at diagnosis continues to change with an increase in colonic location. Key messages What is already known? It is unclear whether the incidence of IBD has now plateaued in urbanised nations. Changes in Crohn’s disease location are often not reported in incidence studies and terminal ileal disease has usually been reported as the commonest site of disease What is new here? The incidence of UC and Crohn’s is no longer rising in Cardiff UK, but the phenotype has changed progressively over time with a continuing increase in colonic disease location and decrease in isolated terminal ileal disease How can this study help patient care? Understanding that Crohn’s colitis is the predominant location has implications for diagnostic tests and implications for treatment options IMPACT STATEMENT This work shows that although IBD incidence is no longer rising, the pattern of Crohn's disease is changing with more colonic disease and less isolated terminal ileal disease. PRACTITIONER RELEVANCE STATEMENT The changing pattern of Crohn's disease location has implications for diagnostic assessment and treatment of this disease.


Introduction
The incidence and prevalence of inflammatory bowel disease (IBD) is increasing in newly industrialised countries across the world, whilst established industrialised nations are seeing a plateau in incidence with a continuing increase in prevalence [1] due to low mortality of IBD and ageing populations [2]. This stabilisation in incidence with increasing prevalence has been demonstrated in recent studies in the United Kingdom [2][3][4].
In Cardiff, incidence data of Crohn's disease (CD) has been reported since 1934 [5,6] and ulcerative colitis (UC) since 1968 [7]. The incidence of CD rose dramatically between 1950 and the mid-1980s and then rose slowly until 2005 [6].
Historically, Cardiff was well-suited to population-based studies owing to its stable population. More recently there has been population expansion and increasing numbers of students. The population in 2011 was 346,090 [8]. The city of Cardiff is served by two hospitals in the same University Health Board (Cardiff and Vale), and nearly all patients are diagnosed and managed there, as well as significant numbers of patients referred from elsewhere in south Wales. With a wealth of historical data for comparison, an update for incidence is of significant interest. In this study, we present incidence data and phenotypic data for patients diagnosed from 2005-2016 with CD, UC and IBD unclassified (IBD-U) whilst resident in the city of Cardiff.

Methods
A population-based incidence study was performed, collecting patients with a diagnosis of IBD made whilst resident in the city of Cardiff, defined as a residential postcode within the Cardiff local authority area which comprises 29 electoral wards. Patients diagnosed with IBD in the 12 years between 2005 and 2016 inclusive were identified from electronic datasets in Cardiff and Vale University Health Board (C&VUHB): three pathology archives (with searches for gut biopsies with inflammation, or a diagnosis of CD, UC or IBD-U), and the hospital Patient Management System, which records diagnostic codes for all in-patient admissions and day-case procedures, including endoscopy, (searching for ICD10 codes K50-CD, K51-UC or K52.3-IBD-U). An archived ADAM (Fujifilm Europe GmbH) endoscopy database was used for colonoscopy and flexible sigmoidoscopy reports, and electronic records of clinic letters from C&VUHB Clinical Portal and from the Welsh Clinical Portal were used to confirm diagnosis. All patients resident in Cardiff were included if they had evidence of at least one clinical encounter for inflammatory bowel disease. The histology database for 1987-2004, and patient management system (1998)(1999)(2000)(2001)(2002)(2003)(2004)) was used to crosscheck and exclude patients with a diagnosis prior to 2005. Clinical letters and endoscopy reports at diagnosis were also checked to confirm that these were incident cases, with no past diagnosis of IBD. All tertiary referral patients resident elsewhere at diagnosis were excluded. Students resident in Cardiff at time of diagnosis were included, but those resident elsewhere at time of diagnosis were excluded. During the period 2005-2016 all cases where the histology was borderline were reviewed in a weekly gastrointestinal pathology meeting, and diagnosis was based on the Lennard-Jones criteria [9]. Date of diagnosis was defined by the first investigation that demonstrated evidence of IBD. For example, the date of endoscopy determined date of diagnosis rather than when the histology was reported, or when the diagnosis was explained in clinic. In some cases where the diagnosis remained unclear for some time, but were subsequently confirmed as IBD, based on further investigations and follow-up, the diagnosis date was recorded as the time of the initial investigations. The majority of cases included had at least five years follow-up and any change in diagnosis was recorded at five years. Disease behaviour was classified according to the Montreal criteria [10].
Data on incidence of CD and UC from previous published Cardiff studies were available. Methodology for case finding in these previous studies was very similar, using discharge coding, histology records and clinical letters, but prior to 1995, much of this was not stored electronically and required searches of paper case files. Current data were compared to historical data where available. Incidence data for Crohn's disease were available back to the 1930s, but only data from 1945 are presented as prior to this numbers were very small. Previous ulcerative colitis incidence data, however, were only available for the period 1968-1975, and was not collected between 1976 and 2004.
Data were collected and summarised using Microsoft Excel for Office 365 (version 2206). Incidence was corrected to UK population age-sex distribution with 2011 Census data [8]. Categorical differences in clinical features between disease types were assessed using Chi-square analysis (IBM SPSS version 27). Trends in Crohn's disease incidence over time from 1951 to 2016 were explored using a JoinPoint regression model (JoinPoint version 4.9.0.1, May 2022, Statistical Research and Applications Branch, National Cancer Institute, USA) [11]. The analysis was used to determine whether there was any inflection/join point in a given year that represented a significant change in the rate of Crohn's disease incidence (where a temporal trend significantly changes). Trends were expressed as the relative annual percentage change. We limited the JoinPoint analysis to a maximum of three inflection point. The threshold for statistical significance was defined as a p value <.05, suggesting the level at which the slope differed from zero.

Results
We evaluated 3422 records from Cardiff and Vale patient management system 2005-2016, and 4631 histology records (2005-2016) ( Figure 1). There were 5276 records after removal of duplicates; 1239 cases had other diagnoses, including diverticular colitis, microscopic colitis, irritable bowel syndrome, colorectal cancer, ischaemic colitis and infective colitis, and tuberculosis in one patient. There were 991 IBD patients, (337 Crohn's disease (CD), 600 UC, and 54 IBD-U). Annual incidence per 100,000 person years (uncorrected for age and sex) are shown in Figure 2.
Colonoscopy or flexible sigmoidoscopy was the main mode of diagnosis (99% for UC and 64% for CD) as shown in Table 1. 27% of CD patients were diagnosed by radiology. Overall 92% of CD had a histological specimen within one year of diagnosis (from subsequent lower gastro-intestinal endoscopy or surgery). Diagnosis was changed in small numbers of patients by five years (Table 1):-4/304 (1%) CD patients had diagnosis changed to UC, 11/528 (2%) of UC had a change to CD and 3/528 (1%) to IBD-U, and 3/46 (7%) of IBD-U had a change to CD, and 9/46 (20%) to UC. This left a total of 45/991 (5%) with a diagnosis of IBD-U after five years.

Crohn's disease
There were 337 new diagnoses of Crohn's disease in the 12year period 2005-2016, giving an incidence of 8.1 per 10 5 person years (95% CI 7.3-9.0) overall. Incidence corrected for age and sex to the UK population (2011 census data) [8] was 7.7 per 10 5 person years (95% CI 6.9-8.6). The age and sex distributions are shown in Supplementary Figure 1, with 52% females overall. There are more males aged 10-19 than females, but this is reversed for older categories. The median age at diagnosis was 29 (range 5-87). Patients diagnosed at age 0-16 make up 14% of cases.
Compared to incidence rates from previous Cardiff studies [5,6,12,13] the incidence for 2005-2016 is higher than all periods up to 1975. There is a significant difference to the period 1976-1985, but no significant difference to the other time-points after 1975 (Supplementary Table 1). Analysis of the changes over time using JoinPoint regression [11] showed that after a small increase for 1946-1953, there is a significant increase in cases between 1953 and 1980 (annual percent change (APC) 0.18, 95%CI 0.13 to 0.23), followed by a less marked gradient from 1980 to 2016 (APC 0.06; 95%CI À0.02 to 0.14) that is not significantly greater than zero ( Figure 3). The rise in incidence since the 1940s is seen in all Montreal age categories, but most marked in the 17-40 age group (Supplementary Figure 2). The median age at diagnosis has fallen from a peak of 46 (1956-1965) to 29 (2006-2015) (Supplementary Table 2). Extent and behaviour of disease in accordance with Montreal classification are shown in Table 1. Colonic (L2) location is the commonest distribution (42%) followed by ileal disease (L1 29%) and ileocolonic (L3 28%). Two patients had isolated perianal disease. One patient had typical Crohn's fistulae at diagnosis, but no other disease elsewhere until nearly 10 years later when they developed extensive ileocolonic disease. The diagnosis was therefore backdated to the original perianal fistula. The other patient had complex fistulising perianal disease, with at least five EUAs and setons over several years. There were fleshy skin tags and anal fissures at diagnosis, with histology showing granulomatous disease. Four patients had isolated disease above the terminal ileum (L4) with no other gastrointestinal tract involvement. Regardless of other sites affected, 69% had colonic involvement, 57% had ileal involvement, and 12% had perianal disease. Disease location varied considerably with age; however, disease was ileocolonic (L3) in 57% of patients aged 16 at diagnosis, falling to 11% of those over 40. L1 and L2 distribution was lower in 16 s compared to those over 40 (Chi-Sq 41.7, p < 0.001; Supplementary Table  3). Compared to our previous data, there has been a progressive fall in isolated terminal ileal disease and isolated small bowel disease (Figure 4) over the past 50 years. 76% had inflammatory (B1) phenotype at diagnosis, followed by   Table  1, 31% had proctitis, 40% left-sided disease and 29% extensive colitis. In Cardiff UC patients diagnosed between 1968 and 1987 [7], the definitions of extent differed, with distal including proctitis, and also those with sigmoid involvement, making comparison with the current study difficult; however, 54.4% had distal disease, 22.4% leftsided, 23.2% disease proximal to the splenic flexure, so broadly similar.
No incidence data for UC in Cardiff is available for 1988-2004. The incidence for 2005-2016 is higher compared to 1968-1987 ( Figure 5), but it is not possible to determine whether incidence is still rising or has plateaued.

Inflammatory bowel disease unclassified
There were 54 new cases of IBD-U. This is an average annual incidence of 1.3/10 5 /year [95% confidence interval 1.0-1.7]. The median age was 29 (range 6-69), with 52% females. Age and sex corrected data are not shown due to small numbers in this group. Of these patients, 12 (22%) had a change in diagnosis (to CD in 3 and UC in 9 patients).

Discussion
The incidence of both CD and UC in Cardiff during the period 2005-2016 is higher than in previous decades using the same data collection methodology [5][6][7]12,13]. CD or UC incidence during this 12-year period, however, are not increasing significantly.
CD incidence is described in other recent UK studies. CD incidence per 100,000 was 10.7 in Devon (2008-2016) [4], and 13.6 in Lothian (2008-2017 [2]. UK-wide mean annual incidence per 100,000 has been reported as 14.3 (2000-2018) [3] and 10.2 also for 2000-2018 [14]. The latter two studies used the same primary care dataset, but case inclusion criteria differed. Pasvol et al. described geographic variation in incidence across the UK and the incidence in Wales (8.9 per 100,000 person years) [14] was similar to our Cardiff data. The Devon population [4], much closer geographically to Cardiff, also had similar incidence to our age-sex corrected rate of 7.7 per 100,000 person years, compared to the higher rate in Lothian [2]. These groups do not show increasing incidence across their respective study periods as in our study but did not compare with historical data from previous decades. Scandinavian studies vary in demonstrating ongoing increases in incidence. Incidence rates in Sweden peaked in the early 2000s with a small decline since, and a mean annual incidence between 2002-2014 of 9.0 per 100,000 [15]. A Norwegian study showed stable incidence between 2010 and 2017 [16]. Danish registry data shows an increase in incidence of CD between 1995-1998 and 2009-2012. Female incidence rose from 7.8 to 10.3 per 10 5 person years and male incidence rose from 5.6 to 8.9 [17], with similar increases in a different Danish study [18]. It has been suggested that there may be a 'ceiling' for IBD incidence at a certain level [19], so populations with lower rates will continue to have an increase, while those with very high rates will be at a plateau. The published rates seem so heterogeneous that it is difficult to show evidence for this. What is clear is that environmental factors play a key role in triggering IBD in genetically susceptible individuals, with increases related to increasingly urbanised societies. The increase in incidence in those under 16, relative to those over 40 shown in the Cardiff data has been shown in other recent studies [14,20] but our data do not show that this is the main reason for increasing rates, and may relate either to earlier diagnosis, or to exposure to the environmental factors earlier in life.
The phenotype of CD has changed over time in the Cardiff population, with a gradual increase in colonic disease (L2 or L3) since the late 1960s, and a corresponding fall in isolated terminal ileal disease (L1). Disease location tends to be stable during the course of the disease in an individual [21] (in contrast to disease behaviour), and location has significant genetic determinants [22]. These are exceedingly unlikely to have changed over a few decades, implying that environmental factors account for the changes. In Asian epidemiology studies it has been noted that ileocolonic disease is the commonest location, occurring in over 50% [23]. Other studies have shown a rise in colonic disease [24,25] and ileocolonic [26] location at diagnosis over time up to 2007. The changing pattern of Crohn's location remains unexplained, but is important as disease location may be a significant determinant of response to drug therapy [27]. Other phenotypic data at diagnosis were not systematically collected in previous Cardiff incidence studies, but data from the population-based study of patients diagnosed 1986-2003, show no significant trend in frequency of either perianal disease, nor in stricturing (B2) or penetrating disease (B3) at diagnosis [28].
The frequency of smoking at diagnosis (30%) has fallen since the rate noted in the previous population-based Cardiff study [28] which was 49% for 1986-1991, 39% for 1992-1997, and 41% for 1998-2003. ONS data record a current smoking rate in population aged over 18 in Wales of 22.1% in 2010, falling slightly to 19.2% in 2014 [29]. Our data show higher rates in CD and lower rates in UC, as expected.
The incidence of UC in Cardiff has increased greatly since 1987, but the gap between 1987 and 2005 makes interpretation difficult. Mean annual incidence per 100,000 person years was 6.3 from 1968 to 1987 [7], which has increased to an incidence of 14.4 between 2005 and 2016. The UC incidence per 100,000 person years was 15.4 in Devon [4], and 19.8 in Lothian (2008-2017) [2]. UK-wide studies reported incidence per 100,000 person years of 23.2 (2000-2018) [3] and 15.7 (2000-2018) [14]. The regional incidence for Wales in the latter study was reported as 14.1, similar to our study. Sweden has seen similar change in incidence in UC as for CD, with a peak of UC incidence in the early 2000s then a decline until 2014, and a mean annual incidence of 17.9 per 10 5 person years between 2002 and 2014. In contrast another Swedish study (covering 1963-2010) showed the incidence continued rising between 2001 and 2010 [30]. Two Danish studies show increase in UC incidence from 1995 to 2012 [17] and from 1980 to 2013 [18].
Our study shows that CD at diagnosis has more systemic effects than UC, with nearly half of patients being anaemic (compared to 30% of UC), and more likelihood of raised Creactive protein or low albumin. Rates of hospitalisation at diagnosis were also higher. It is rare for studies to report whether diagnosis was made as an in-patient. Our figure of 20% and 15% and diagnosed as an in-patient for CD and UC respectively are comparable to other reports with hospitalisation for CD approximately 23% [31] and for UC (within one year of diagnosis) 17-29% [32].
The strength of our study is that we have detailed information about clinical features of all our patients included in the study. Because overall case numbers is limited, confidence limits for changes in incidence over time are wide. In our previous studies we have asked general practitioners in Cardiff to check whether they have IBD patients that may not appear in our data. The yield of this step has previously been extremely small (only 2/212 (0.9%) cases [6]). It is plausible that students residing in Cardiff could have had their diagnosis made in their home town, and not been seen in an IBD clinic in Cardiff. Likewise patients studying elsewhere may have been erroneously included in our dataset if they were diagnosed in Cardiff and their student status not recorded. The net effect on incidence would be insignificant.

Conclusion
CD and UC incidence for 2005-2012 in Cardiff is higher compared to previous decades, but the slight increase noted during the last 12 years is not significant. The incidence of Crohn's disease has not increased significantly since the early 1980s. Crohn's disease phenotype continues to change with increasing colonic involvement and less isolated small bowel disease.