Prevalence and long‐term outcome of sub‐clinical primary sclerosing cholangitis in patients with ulcerative colitis

Primary sclerosing cholangitis (PSC) is closely associated with inflammatory bowel disease, particularly ulcerative colitis (UC), with an increased risk of biliary and colorectal malignancy. We sought to clarify the prevalence, characteristics and long‐term outcome of sub‐clinical PSC diagnosed by magnetic resonance cholangiogram (MRC) in patients with UC and normal liver biochemistry, with or without colorectal dysplasia (CRD).


| INTRODUC TI ON
Primary sclerosing cholangitis (PSC) is a progressive liver disease characterised by inflammation and fibrosis leading to multi-focal biliary strictures and cirrhosis. 1 The clinical presentation of PSC is variable; 50% of patients are asymptomatic at presentation and identified while investigating cholestatic liver biochemistry. 2 An elevated alkaline phosphatase (ALP) is the most frequent abnormality, although this can be normal in approximately 8.5% of cases, and may fluctuate during the course of disease. 3 The diagnosis of large-duct PSC is established on the basis of a cholangiogram. Magnetic resonance cholangiography (MRC) is now considered the gold standard in diagnosis, as it is non-invasive, uses non-ionising radiation and can delineate distal bile duct anatomy in the presence of proximal obstruction. 4 A position paper from the international PSC study group (IPSCG) confirms the importance of MR imaging in the assessment of PSC. 5 In a meta-analysis of 456 subjects (185 with PSC and cholestasis), MRC was reported to have a high sensitivity (86%) and specificity (94%) for detecting PSC. 6 PSC is strongly associated with inflammatory bowel disease (IBD), with a reported association in Northern Europe and America of 60-80%. 7 Ulcerative colitis (UC) accounts for most cases and often precedes the development of PSC, although it may be diagnosed at any time during the course of the liver disease. 8 Traditionally, PSC is diagnosed in 0.8% to 5.6% of patients with UC, more frequently in those with extensive (5.5%) rather than distal (0.5%) colitis. 9 A recent Norwegian study of 322 IBD patients screened with MRC found 8% had diagnostic features of PSC. 10 Only 2.2% had been previously known to have PSC. 10 The colitis associated with PSC has characteristic features: pancolitis with co-existent ileitis ("back-wash ileitis"), more severe inflammation in the proximal colon, non-inflamed rectum ("rectal sparing"), and a quiescent or prolonged sub-clinical course, compared to UC without PSC. 10 Genome-wide association and susceptibility studies suggest that PSC-UC is a unique disorder, genetically distinct from classical UC. 11,12 Patients with PSC-UC are at a greater risk of developing colorectal carcinoma (CRC) and CRD than those with UC alone; a meta-analysis of 11 studies concluded that the risk was increased by approximately four-fold. 13 In view of this high risk of colorectal malignancy, and in contrast to classical UC without PSC, current guidelines recommend annual surveillance colonoscopy in patients with PSC-UC, starting from the time that PSC is diagnosed. 14 Patients with PSC are also at an increased risk of cholangiocarcinoma and gallbladder carcinoma 15 ; and annual surveillance by abdominal ultrasound is recommended to detect gallbladder polyps, which may be premalignant. 14 Given the distinct phenotype and increased risk of malignancy in patients with PSC-UC compared to UC alone, we sought to determine whether sub-clinical PSC (defined as radiological PSC in the presence of normal liver biochemistry) could be detected by MRC in patients with UC and no clinical or biochemical evidence of liver disease. We aimed to investigate the prevalence and clinical course of PSC in two cohorts: patients with extensive UC and normal liver biochemistry, and patients with UC complicated by CRD and normal liver biochemistry. We hypothesised that those patients with UC-CRD may have a higher prevalence of sub-clinical PSC than those with UC alone. We followed these cohorts with blood tests, imaging and endoscopic surveillance as per current guidelines, to establish whether there was progression of disease and/or development of complications.

| Extensive UC Cohort
Patients with UC were prospectively recruited (sequential patients given the opportunity to opt in) from the IBD clinic at the John Radcliffe Hospital in Oxford from 2005 to 2008.

Lay Summary
We performed magnetic resonance bile duct imaging in patients with ulcerative colitis who had normal liver blood tests, with and without a prior history of colon dysplasia/ cancer. We demonstrated unsuspected bile duct abnormalities in 21% and 14% of these patients respectively.
We then followed them up over 10 years, with one-third developing abnormal liver biochemistry, one-fifth developing progressive bile duct disease and over half developing malignancy. We suggest consideration of bile duct imaging in those with ulcerative colitis, irrespective of liver function tests.

| CRD cohort
Patients with UC and histologically confirmed CRD/CRC identified at colorectal surveillance endoscopy were recruited from the IBD clinic at the John Radcliffe Hospital in Oxford from 2007 to 2010.
The diagnosis of CRD/CRC was identified from the histopathology database searched from 1990 to 2010.

| Controls
Positive controls were recruited from the hepatology outpatient clinic with an established diagnosis of PSC, confirmed on MRC (n = 28/28) and liver biopsy (n = 20/28

| CRD cohort
Inclusion criteria were as for the extensive UC cohort, with the addition of histological evidence of CRD or CRC within a segment of colitis. Exclusion criteria were as for the extensive UC cohort with the additional exclusion of those with histologically indefinite CRD or sporadic adenomas arising proximal to the upper limit of colitis.

| Definitions for this study
The extent of colitis was defined as the maximum documented extent histologically at any time during the course of the disease.
Liver biochemistry was defined as normal if the serum alkaline phosphatase (ALP), gamma-glutamyl transferase (GGT) and alanine transferase (ALT) values were below the upper limit of normal for our institution at the time of recruitment and on all previous hospital records. An isolated mildly elevated bilirubin (up to two times the upper limit of normal) was accepted after assessment by a consultant hepatologist (RWC) as attributable to Gilbert's syndrome. The presence of CRD or CRC was identified at colonoscopy and confirmed histologically. CRD was classified at the time of this study, as either flat or associated with a raised lesion or mass (dysplasia-associated lesion or mass (DALM) or adenoma-like mass (ALM)), and confirmed histologically as arising within an area of documented colitis. In accordance with current SCENIC classification, these would be classified as endoscopically invisible dysplasia, endoscopically resectable non-polypoid and polypoid lesions. 16 Two separate gastrointestinal pathologists blinded to clinical details (BW and DD) reviewed the colonic biopsies and re-defined the degree of dysplasia using the WHO classification: (a) low-grade CRD (none, mild or moderate); or (b) high-grade CRD (severe dysplasia, carcinoma in situ or intramucosal carcinoma). 17

| Enrolment of patients and controls to the extensive UC and CRD cohorts
Enrolment to both cohorts of this study for cases and controls is outlined in the flowcharts in Figure S1A and B.

| Extensive UC cohort
Sixty-two patients identified from the IBD clinic with documented extensive UC and normal liver function tests opted into this study.
Eleven of these were excluded; five because of limited extent or incorrect classification of colitis on review, five because of claustrophobia or other contraindication to MRC and one because of abnormal liver biochemistry at recruitment itself. Fifty-one patients participated, who all underwent an MRC and blood tests at recruitment. Forty-eight patients were followed up in IBD clinic in Oxford and three patients were follow-up at other UK institutions, contacted by telephone to establish clinical outcome.
Routine follow-up included annual review with repeat blood tests, with abdominal imaging and endoscopic surveillance as clinically indicated according to national guidelines. Demographics, clinical details, laboratory data, endoscopy and histology reports were reviewed retrospectively.
Twenty-eight healthy volunteers (negative controls) were matched for age and gender, and underwent an MRC under the same conditions as the cases. Twenty-eight patients with PSC (positive controls) were matched for age and gender, and their MRC scans and blood tests at the time of diagnosis were retrospectively reviewed.

| CRD cohort
Forty-eight patients were identified from the histopathology database with UC and histological evidence of CRD. Twenty-eight patients who attended the IBD clinic with UC and prior CRD/CRC opted into this study. Eight were excluded; three because of the presence of sporadic adenomas, two because of incorrect classification of colitis, two because of contraindications to MRC and one because of distance living away from the hospital. Nineteen patients participated, who all had an MRC and blood tests at recruitment.
At the analysis stage, two patients were excluded from this study because consensus histological opinion defined "indefinite" rather than low-grade dysplasia. Thus, 17 patients received follow-up in IBD clinic (virtually and/or in person); 15 in Oxford and 2 elsewhere who were contacted, as per the extensive UC cohort patients, to establish clinical outcome. MRC scans from 24 positive and negative controls (age and gender matched) were blinded and re-analysed for the CRD cohort.

| Ethics and consent
Ethical approval was granted from the Oxfordshire Research Ethics Committee (Extensive UC cohort: 05/Q1606/151 and CRD cohort: 09/H0607/4). Written informed consent was obtained from each patient and control for this study. A letter was sent to the patient/ control with a copy to their GP informing them of the results of their study bloods and MRC scan. Results that required further discussions were followed up by the research team and contact made. Further assessment was determined by clinical need. For extensive UC patients this was routine in outpatients with liver tests performed on an annual basis and colonoscopy as per disease duration and severity.

| Evaluation of MRC scans for biliary changes consistent with PSC
Each MRC was evaluated for evidence of intrahepatic duct or extrahepatic duct irregularity, beading, strictures, associated biliary dilatation, diverticula or webs, an increased angle of ductal confluence and/or "disconnected" peripheral ducts. The certainty of the radiologist's interpretation was scored through a three-level classification of "definite," "possible" or "absent" biliary abnormalities consistent with PSC, according to Gestalt theory. 18,19 The final diagnosis made was based on separate expert opinions of two Hepatobiliary Radiologists, and where there was disagreement consensus was reached.
A classification of "definite" was made when clear irregularity, beading and strictures of the ducts were visible. A classification of "possible" was made when the ducts (usually intrahepatic) were irregular, with a degree of uncertainty as to whether this was a true finding, caused by blood vessels impinging on the ducts, or artefact caused by technical factors. "Absent" biliary abnormalities was defined by the absence of biliary irregularity, strictures and/or dilatation. The radiological diagnosis of cirrhosis was made by assessing regional changes in hepatic morphology.

| Blinded interpretation of MRC scans
MRC scans were individually interpreted by a gastrointestinal radiologist in the study team within 24 hours of conducting the scan and an individual report prepared for each scan. Subsequently sheets. Both were blinded to the clinical profiles and any previous imaging, to avoid possible bias. All scans included in this study were deemed of adequate quality to identify evidence of intra-or extrahepatic duct abnormalities.

| Intra-and Inter-observer variation
Methods for MRC pooling and reporting are detailed in Figure   S2. All MRC images of the cases and controls were pooled (total 107 MRCP scans for extensive UC and 67 MRCP scans for CRD cohorts) and reported on two separate occasions (total 214 independent reads for the extensive UC cohort) by two consultant gastrointestinal radiologists with a special interest in hepatobiliary disease (HB with 10 years' and MB with 9 years' consultant experience at the time of reporting). In the event of intra-and inter-observer disagreement, definitive decisions were taken by consensus between the two radiologists. The overall quality of MRCs was considered high, with only one study repeated because of poor quality.

| Statistical analysis
Cases were divided into two groups according to the presence or ab- IgG4 P = .79).

| Prevalence of Sub-clinical PSC
MRC interpretation is shown in

| Prevalence of Sub-clinical PSC after pooling and blinding of MRC scans
MRC interpretation after pooling of randomised and blinded scans is shown in   Table S2.

| MRC Inter-and Intra-observer variation
MRC observer variation after pooling of blinded scans is detailed in Table S3. Inter-observer agreement between radiologists for all scans Intra-observer agreement by radiologists for MRC interpretation was defined as almost perfect (kappa = 0.96). Overall, 100% of positive PSC controls and 98% of negative healthy controls were identified correctly.

| MRC features in those with Sub-clinical PSC
MRC evidence of biliary abnormalities consistent with PSC for the two cohorts is shown in the flowchart in Figure 1A and B.
Representative MRC images of 'definite' and 'possible' PSC are shown in Figure 2A

| Variables associated with Sub-clinical PSC
Demographics and characteristics of UC patients with sub-clinical PSC are shown in Table S5. Variables associated sub-clinical PSC in the two cohorts are shown in Tables 3 and 4 (Table 4).

| Outcomes in sub-clinical PSC
Patients with UC were followed up in the IBD clinic and underwent colonoscopic surveillance; the frequency was determined by activ-

| Comparison of outcomes in cases and control groups
Clinical outcome during long-term follow-up of all cases and controls is shown in Table S7. Cirrhosis (P = .03) and hepatobiliary car-

| D ISCUSS I ON
In this prospective study, the prevalence of sub-clinical PSC on MRC was two-to three-fold higher (14%) in those with extensive UC and normal liver biochemistry, and three-to four-fold higher (24%) in those with UC and prior CRD/CRC, compared with the reported prevalence in those with UC and cholestatic liver biochemistry (0.8%-5.6%). 9  In the Norwegian sub-clinical PSC cohort, the patients were older and there were more females compared with classical cohorts, and this was also the case in our cohort. 10  The radiological diagnosis of sub-clinical PSC was considered definite in 5/11 cases and none of the negative controls. All those with biliary abnormalities on MRC had IHD irregularity, which suggests that in asymptomatic patients with UC and normal liver function, IHD irregularity is the first manifestation of PSC.
The accuracy of MRC in the diagnosis of PSC may be influenced by radiological experience and disease spectrum. 33 In this study, we used a three-tier definition of definite, possible or absent PSC to define biliary abnormalities consistent with PSC, as used previously, 34 with a higher yield to detect more subtle biliary abnormalities when compared to alternative definitions based on duct stricturing. 32 Our inter-observer and intra-observer variations were small.
Inter-observer agreement for intrahepatic bile duct abnormalities was better than for extrahepatic duct disease, mirroring a study by Vitellas et al 4 The optimal protocol for MRC has only recently been defined, 5 with differences in instrument manufacture or magnetic field strengths accounting for variability between studies. In this study, the same MRC scanner was used for all cases and negative controls, although some of the positive controls were imaged at an earlier date. The overall quality of MRCs was high, with only 1 of 174 MRC scans needing to be repeated. 24 Patients with classical PSC-UC are at a four-fold greater risk of developing CRD or CRC than those with UC alone. 35,36 Because of the increased CRC risk, patients with PSC-UC are enrolled in surveillance programs with annual colonoscopy and biopsies from the time of diagnosis of PSC. 14 Importantly, we detected a three-to four-fold increased frequency of sub-clinical PSC in patients with UC and a prior history of CRD/CRC compared to that described for classical PSC. Furthermore, all these patients had been diagnosed with colorectal adenocarcinoma or HGD lesions, not LGD lesions, three of whom required colectomy and one who developed a new adenocarcinoma in the cuff 17 years after pouch surgery for CRD.
This finding is supported by evidence that PSC-IBD cancers are more aggressive than IBD cancers at diagnosis. 37 Interestingly, elevated serum IgG4 levels were found more frequently in the group with HGD or CRC. IgG4 has been described as an immune mediator in the context of malignancy, 38 and an abundance of mucosal IgG4 in UC has been associated with more severe disease activity and the presence of PSC. 39 Early identification of sub-clinical PSC, with an increased frequency of colorectal surveillance, may have had an impact on CRD detection and management. The role of UDCA in chemoprevention of CRC remains debated. 40,41 The main limitations of this study are (a) its small sample size,