Surveillance for cholangiocarcinoma in patients with primary sclerosing cholangitis: Can we be more proactive?

The standardised mortality ratio of primary sclerosing cholangitis (PSC) is 3.55 (2.94–4.28)¹ with most excess deaths due to malignancy. Patients with PSC often express fear of developing cancer.² For many patients, this is the most difficult aspect of PSC.^{3 4}

68%–80% of patients with PSC develop inflammatory bowel disease (IBD),^5–8 which confers an increased risk of colorectal cancer (CRC). Thus, annual surveillance colonoscopy is recommended in UK, American and European Guidelines.^9–11 This is alongside 6–12 monthly ultrasound surveillance for hepatocellular carcinoma and to detect potentially malignant gallbladder polyps.¹² Some studies also report an increased incidence of pancreatic cancer,^{1 5} although inconsistently.¹²

However, most cancer-related deaths in PSC result from cholangiocarcinoma (CCA).6 7 13–16 The overall annual incidence of CCA in PSC is 1.1 (0.5–2.7)%¹⁷; 160–1600 times that of the general population.1 5 6 12 18 Up to 60% of patients with PSC with CCA die within a year of diagnosis,^{6 15} and despite resection, 5-year survival is ≤40% (20% with tumours >3 cm).¹⁹ Thus, CCA accounts for 24%–58% of deaths in PSC,6 7 13 16 which exceeds combined deaths attributable to CRC (2%–11%),^{6 13 16} gallbladder cancer7 12 13 15 and hepatocellular carcinoma.1 7 12 15 Despite these poor outcomes, guidance on surveillance for CCA is rather tentative (table 1), often with a ‘suggestion’ or ‘conditional’ recommendation for annual imaging and weak and inconsistent recommendations for biomarker monitoring.

Table 1.

Current recommendations from guidelines and expert review groups²⁴ regarding cholangiocarcinoma surveillance in PSC

European Association for Study of the Liver 202211	Surveillance with ultrasound and/or MRI/MRCP for cholangiocarcinoma and gallbladder malignancy is suggested at least yearly in patients with large-duct disease, regardless of disease stage. o.	Not suggested for surveillance purposes due to its insufficient accuracy.
American Association for Study of Liver Disease 201010	In the absence of evidence … many clinicians screen patients with an imaging study at annual intervals. Recommend annual ultrasound to detect mass lesions in the gallbladder.	As for imaging
American College of Gastroenterology 201544	Consider screening for CCA with regular cross-sectional imaging with ultrasound or MR every 6–12 months. (Conditional recommendation, very low quality of evidence)	As for imaging
Schramm et al 24	There is no quality evidence supporting or refuting CCA screening. However, many experts …recommend regular CCA screening with MRI/MRCP.	Not discussed
American Gastroenterological Association 201945	Should include imaging by ultrasound, CT or MRI every 6–12 months	Optional add-on to imaging
British Society of Gastroenterology 20199	Suggest gallbladder scan annually for polyps. Lack of clear data in asymptomatic patients	Not recommended. little evidence to justify
	Imaging	Serum Ca.19.9

CCA, cholangiocarcinoma; PSC, primary sclerosing cholangitis.

Possibly reflecting this equivocation, clinical practice varies. In a UK multicentre audit²⁰ of 1795 patients with PSC (90% cared for by a hepatologist), 71% were reported as undergoing annual hepatobiliary imaging: most commonly by ultrasound (47%) or alternating ultrasound and MRI (36%). Frequency of imaging actually performed was not assessed but was recently reported in a single-centre audit,²¹ obtaining similar percentages. In contrast, in recent European surveys,^{22 23} 90% of centres reported performing routine imaging surveillance for CCA, most often by annual MRI, despite absence of firm recommendations in UK⁹ and European¹¹ guidelines.

Expert groups have stated a lack of evidence as the reason for not providing more definitive advice regarding surveillance in clinical practice.²⁴

Additionally, 27%–38% of CCAs present within the first year after PSC diagnosis,5–7 12 18 25 with incidence then falling: in four studies^{5–7 25} from 2.9 (1.9–4.0)% during the first year to 0.76 (0.72–1.20)% over subsequent years. Recently, a prospective Swedish study (512 prevalent patients) found an unusually low annual CCA incidence of only 0.43%.²⁶ This has implications for the efficacy of a surveillance programme, with the aims of achieving earlier cancer detection and improved survival.

However, the evidence base for cancer surveillance protocols is often incomplete. The incidence of CRC is also highest early in the course of PSC-IBD,^{15 25} yet indefinite annual colonoscopic surveillance is accepted, despite absence of supportive randomised controlled trial data. Instead, there is only observational evidence in IBD, with ⁶ or without²⁷ PSC, wherein lesions diagnosed by surveillance are (compared with CRCs presenting clinically) more often resectable and associated with better postcancer survival.

Comparable evidence is accumulating regarding CCA surveillance in PSC. Ali et al ²⁸ reported on a hepatobiliary cancer surveillance programme at the Mayo Clinic, comprising annual imaging (ultrasound or MRI) and serum Ca-19.9. Of 890 patients with PSC, 79 developed hepatobiliary cancers (54 CCA). Tumours detected by the surveillance programme (n=40) were of a less advanced stage compared with the 39 detected clinically, more likely to undergo curative treatment, and resulted in greater 5–10-year survival. The sustained difference in survival argues against this being simply a lead-time bias effect. Serum Ca-19.9 values were elevated in 6/54 patients with CCA despite initially normal imaging, indicating that this also contributed to CCA detection as part of a surveillance (as opposed to diagnostic) programme. Diagnostic accuracy of serum Ca-19.9 in this regard has been questioned⁹ but may be enhanced by serial measurements over time.²⁹

In a follow-up study, Eaton et al ³⁰ compared the accuracy of contrast MRI and USA, each performed within 3 months of one other, and reviewed blindly, among 266 patients with PSC (146 without CCA and 120 with early (≤3 cm) perihilar CCA). MRI was found to be only marginally less specific (86% vs 90%) but more sensitive (89% vs 52%) than ultrasound for detecting possible or definite perihilar CCA, with a sensitivity advantage also in asymptomatic patients (71% vs 29%). Moreover, in asymptomatic patients, perihilar CCAs diagnosed on MRI were smaller, less advanced and had higher 5-year survival (100% vs 33%) and lower recurrence rate than those also detected on ultrasound. A similar trend was found in asymptomatic patients with CCA developing more than 1 year after PSC diagnosis. Although rigorously performed, this study has limitations (retrospective, non-randomised, gold standard possibly favouring MRI). Furthermore, contrast enhancement with gadolinium was used; this may enhance sensitivity,²⁴ but repeated use may cause brain deposition of gadolinium (although of unknown clinical significance).³¹ Importantly, the sensitivity of non-contrast magnetic resonance cholangio-pancreatography (MRCP) is approximately 10% lower than that of contrast-enhanced MRI.³²

Recent observational studies from the UK (n=2 588 PSC)¹⁵ and from a European survey²³ (n=2975 PSC, 28 centres) also suggest that patients undergoing annual imaging had lower risks of death, respectively, from hepatopancreatobiliary cancer (80% CCA, hazard ratio: 0.43)¹⁵ and from any cause (hazard ratio: 0.53).²³

While other evidence also suggests that MRI is more ‘accurate’ than ultrasound in detecting small hepatobiliary cancers,³³ arguments against routine MRI-based surveillance include the potential risks of repeated contrast exposure, cost implications and limited availability of hepato-pancreato-biliary (HPB) radiology expertise. Moreover, the positive predictive values reported by Eaton et al ³⁰ depends on the ‘true’ annual incidence of CCA. Assuming the median incidence value cited above of 2.9% (over the first year), 0.76% (over subsequent years) and 0.43% (among prevalent PSC patients in the Swedish study)²⁶ yields PPV values for suggestive MRI CCA appearances of 16%, 4.6% and 2.7%, respectively. Thus, for every true CCA case diagnosed by a surveillance programme, there would be 5, 21 and 36 patients, respectively, with false-positive scans. Such patients may require invasive procedures such as ERC, cholangioscopy and even resection, purely for diagnosis. These interventions are relatively high-risk in PSC and are not always definitive, generating patient anxiety and further workload for clinicians. This is especially problematic in the UK, with a radiologist workforce that by 2025 will be 39% below requirements.^{34 35}

Is there a role for endoscopic cholangiography (ERC) in CCA surveillance? Current UK practice is an ‘on-demand’ ERC strategy, for example, in patients developing a dominant stricture. ERC in PSC should be undertaken by experienced endoscopists after discussion within a multidisciplinary team (MDT). A German centre has, however, reported an association of routine ‘scheduled’ (compared with ‘on-demand’) ERC in PSC, with improvement in transplant-free survival (17.9 years vs 15.2 years, p=0.008).³⁶ Results from a European multicentre survey²³ also suggest reduced mortality rates in patients undergoing ERC surveillance, compared with only MRI or US surveillance (8.5 (6.3–10.5) vs 12.5 (10.6–14.5)%). Furthermore, digital single-operator cholangioscopy has improved direct visualisation of mucosal abnormalities and hence diagnostic yield as well as enabling application of newer diagnostic technologies.³⁷

Given these developments, it may not be in the interests of patients with PSC, for clinicians simply to accept current practice and await confirmation of observational cohort data.^{23 28 30} Or to await validation of better diagnostic tools for CCA in PSC.³⁸ A PSC support patient survey (n=190)⁴ found that 84%–90% considered it ‘extremely important’ to improve cancer screening for people with PSC, and to ‘look for new ways to detect cancer’.

A randomised trial of routine surveillance versus none might be unfeasible given the likely preferment of informed patients for surveillance (as demonstrated for HCC in cirrhosis,)³⁹ with reluctance to accept randomisation; and also, the potentially low threshold for clinician-driven imaging in the non-surveyed group. Instead, we propose consideration of:

1. retrospective/prospective audit in the UK-wide PSC cohort,²⁰ of (1) surveillance investigations performed year-on-year (collected electronically from participating hospitals), (2) (review of their diagnostic accuracy, by (consent-driven) centralised and blinded review and how accuracy relates to size, resectability and outcome of CCAs diagnosed, (3) health-economic evaluation of ensuing investigations to resolve suggestive findings, (4) evaluating patients’ views on, and understanding of, the goals of CCA surveillance.

2. Based on European practice and on the Mayo Clinic evidence cited above, updating the UK-PSC/BSG Guidelines to recommend more explicitly 6–12 monthly routine CCA surveillance (potentially incorporating serum Ca-19.9 and MRI/MRCP)

3. A national certification programme for radiologists reporting MRCPs in patients with PSC, including (1) imaging review in experienced centres, incorporating an MDT and with access to clinical trials,²⁴ (2) establishing standards of practice akin to CT Colonography reporting for the NHS Bowel Cancer Screening Programme.⁴⁰ Radiologists could receive specific training, including supervised interpretation of validated cases.

4. Personalised use of tumour biomarkers for serial monitoring according to underlying genotype. Genotype-dependent cut-offs of Ca-19–9 and CEA result in a 43% reduction in false-positive cancer diagnoses, with genetic variants of fucosyltransferase (Fut)-2 and Fut-3 polymorphisms.^{41 42}

5. Developing potentially more effective CCA surveillance strategies, including (1) annual gadolinium-enhanced MRI and (2) a scheduled annual (vs on-demand) ERC strategy in patients with dominant strictures. Evaluation of such strategies might be best performed as substudies of large multicentre RCTs evaluating novel therapeutic or prophylactic drug therapies in PSC. Potential drugs could include statins,⁴³ aspirin and immunomodulatory agents.

Such measures would hopefully result in a more proactive (and potentially survival-enhancing) strategy for CCA surveillance in PSC.

Ethics statements

Patient consent for publication

Not applicable.

Ethics approval

Not applicable.