Diagnosis of functional strictures in patients with primary sclerosing cholangitis using hepatobiliary contrast-enhanced MRI: a proof-of-concept study

Primary sclerosing cholangitis (PSC) is a chronic, idiopathic, fibroinflammatory cholestatic liver disease that causes strictures of the intra- and/or extrahepatic biliary tree leading to progressive biliary and hepatic damage that results in portal hypertension and hepatic dysfunction (HD) within 10 to 15 years after the initial diagnosis [1]. Orthotopic liver transplantation (OLT) is the only effective treatment [1]. PSC patients are also prone to develop cholangiocarcinoma (CCA), gallbladder carcinoma, and, in the setting of cirrhosis, hepatocellular carcinoma (HCC). Furthermore, those with concurrent IBD are at risk of colorectal cancer [2]. Because half of PSC patients will develop a so-called dominant stricture (DS) over the course of the disease, it is imperative to diagnose such a stricture early, and to determine its location, as well as its clinical significance, to optimize management and render a prognosis [3, 4].

Both the American Association for the Study of Liver Diseases (AASLD) and the European Association for the Study of the Liver (EASL) guidelines define a DS as a lumen diameter ≤ 1.5 mm in the common bile duct (CBD) or ≤ 1 mm in either the main right (RHD) or left hepatic duct (LHD) within 2 cm of the hilum, as measured on endoscopic retrograde cholangiopancreatography (ERCP) [5, 6]. However, applying these cutoffs to conventional T2-weighted magnetic resonance cholangiopancreatography (T2-MRCP) [7] has been unsuccessful due to technical differences in image acquisition [8]. To address this problem, stricture severity, defined as either ≥ 75% (high-grade stricture, HGS) or < 75% (low-grade) narrowing of the CBD or hepatic duct lumen [8, 9], was introduced specifically for T2-MRCP. Even so, despite being defined for T2-MRCP, the morphologic appearance of HGS provides no more information about the clinical relevance of a stricture than DS [4, 7, 8]. However, it remains to be seen if the term HGS, which was accepted into the PSC imaging lexicon in 2021, will be widely adopted by radiologists and referring physicians [8].

Thus, recently, the consensus opinion of the International PSC Study Group defined a DS as possible if a segment of the extrahepatic or first-order intrahepatic ducts is narrowed on high-quality T2-MRCP or ERCP with either: (a) a 2-month history of worsening cholestatic symptoms plus recent, elevated bilirubin and/or alkaline phosphatase level(s) 1.2 × above baseline; or (b) elevated bilirubin and/or alkaline phosphatase level(s) 1.5 × above baseline within the previous 6 months [10]. However, this is a very controversial point. According to the recent MR Working Group of the International Primary Sclerosing Cholangitis Study Group, the term DS should not be used with MRCP. The study group recommends using the term HGS, which refers to the morphology of the duct(s) [8]. Regardless, the triad of clinical, laboratory, and radiologic features, although suggestive [7, 8], precludes confident, accurate diagnosis that would enable timely intervention to avert OLT and/or mortality.

To date, no universally accepted nomenclature defines a diagnostic biliary stricture of potential prognostic relevance. An ideal definition would be based on an objective, easily reproducible feature with high inter-reader agreement. Therefore, we propose the term “potential functional stricture” (PFS), defined as impaired gadoxetic acid (GA) excretion on 20-min, T1-weighted hepatobiliary contrast-enhanced MR cholangiography (T1-MRC). GA (Primovist® in Europe, Eovist® in the USA), normally taken up by functioning hepatocytes, then excreted via the bile ducts and kidneys within 20 min, allows visualization of physiologic bile excretion [11]. GA-MRI, including T1-MRC, has shown high sensitivity and specificity in diagnosing various biliary disorders, e.g., PSC and post-OLT anastomotic strictures [11‐14].

We chose the term “potential functional stricture” (PFS) consciously, as impaired gadoxetic acid excretion can be due to either a true functional stricture (FS), i.e., significant bile duct narrowing with mechanical blockage or to hepatocellular dysfunction (HD) that is characteristic of advanced PSC. If “potential functional stricture” (PFS) is diagnosed, T2-MRCP, in addition to all-sequence MR, can help distinguish significant biliary obstruction, i.e., true FS, from HD of advanced PSC, as previously reported [15, 16]. A pruned tree appearance [17] and features of liver cirrhosis and/or portal hypertension are consistent with HD, once dilated bile ducts due to biliary obstruction can be excluded, as previously described by the functional liver imaging score [18‐20]. In ambiguous cases or if clinically indicated, ERCP can be additionally performed.

Therefore, by determining if excretion is normal (i.e., no functional stricture (NFS)) or impaired (i.e., PFS) on the HBP of GA-MRI, using this binary system, we had two aims: to assess inter-reader agreement and prognostic value of PFS diagnosis and location on T1-MRC versus DS or HGS diagnosis and location on T2-MRCP in PSC patients, using ERCP as the gold standard.

Spleen volumes were significantly higher (p = 0.016) with impaired excretion (PFS), and significantly more patients with splenomegaly had liver-related events (p < 0.001) (Table 2 and Table 1S).

Our suggested binary stratification into normal (NFS) versus impaired excretion (PFS) on 20-min hepatobiliary phase (T1-MRC) images had 100% positive predictive value. In other words, all 42 PFS patients had either a functional stricture or hepatic dysfunction. ERCP confirmed all but one functional stricture which was later confirmed histologically at OLT. Similarly, the diagnosis of NFS had a 100% negative predictive value. Although the experts had higher inter-reader agreement than the residents, i.e., almost perfect vs substantial, on T1-MRC, this was far better than T2-MRCP agreement for DS and HGS, which had only poor and moderate agreement, respectively. Even using the median T1-MRC diagnoses of all 6 readers, differences between PFS and NFS were still significant for events. Further confirmation of the robustness of PFS over DS and HGS is the relatively few cases with discrepant diagnoses between residents and experts when interpreting T1-MRC (3%) versus T2-MRCP (16–26%).

NFS versus PFS diagnosis correlated far better with clinical outcomes (HR 19.5 vs 2.1 and 1.9) than DS or HGS diagnosis on conventional T2-MRCP, and there was far better agreement on stricture location. Because clinically significant strictures are so common, early and accurate stricture diagnosis is imperative. With prompt dilatation and/or short-term stenting, liver failure and OLT can be postponed [37, 38].

To the best of our knowledge, this is the first MR imaging study based upon Erlinger’s 1985 concept of bile physiology. However, in our definition of PFS, we consider the right and left hepatobiliary tree as two separate systems. Therefore, Erlinger’s definition, i.e., bile’s failure to reach the duodenum, only applies to PFS when a significant stricture, i.e., FS, occurs in the hilum or CBD, or in the setting of hepatocellular dysfunction (HD) [39]. However, in PFS, if there is a blockage of the right hepatic duct causing failure of the right liver lobe, bile (i.e., contrast agent) would still reach the duodenum via the patent left hepatic duct, and vice versa. Determining whether excretion on the HBP of GA-MRI, like bile flow, is normal versus impaired in the presence of significant bile duct narrowing has improved both sensitivity and reader confidence in the diagnosis and location of biliary obstruction [13].

We compared the inter-observer variability between DS and HGS based on T2-MRCP, and our proposed potential functional stricture (PFS) diagnosis on T1-MRC, finding that T1-MRC diagnosis is clearly superior to all T2-MRCP–based scores. We found only moderate and fair agreement for T2-MRCP–based HGS and DS, respectively. T2-MRCP’s poor performance, even among PSC experts, is likely due to the subjectivity in determining bile duct stricture severity [7]. Moreover, applying ERCP criteria to T2-MRCP led to poor anatomic depiction, lack of functional information, and/or an underdistended biliary tree on T2-MRCP contributing to both false-positive and false-negative stricture misdiagnoses [7, 25, 40]. In contrast, there was almost perfect inter-reader agreement among experienced readers, and substantial agreement, even among residents, for both PFS presence and location on T1-MRC.

Beyond easy and accurate stricture diagnosis on T1-MRC, we found a statistically significant correlation with well-established prognostic biomarkers, including the RMRS, ALBI, Short-Term UK-PSC risk score, AOM for PSC score, and, most importantly, primary and secondary clinical endpoints. Our results are in line with previous studies where impaired excretion correlated with elevated liver function tests, Mayo PSC risk scores, and downstream biliary obstruction [41, 42]. Spleen volume also correlated significantly with PSC adverse events, differentiating patients with normal versus impaired excretion. As in other chronic liver diseases, splenomegaly [35] and imaging signs of portal hypertension herald advanced PSC, and increased risk of further event(s) [43].

The Kaplan–Meier curves showed a clear separation between adverse clinical events in T1-MRC NFS vs. PFS subpopulations. Furthermore, multivariate analyses using Cox regression showed that T1-MRC-based diagnosis was an independent risk factor for adverse events in PSC patients, outperforming lab tests and clinical scores, even after categorization into an ordinal system. In addition, except for T1-MRC, no other MR metric correlated with established clinical scores, which was unsurprising, as we know T2-MRCP-DS/HGS-diagnosed patients range from asymptomatic to having abnormal liver function tests and/or cholestasis [4].

The limitation of T1-MRC was its inability to separate impaired excretion patients (PFS) with true stricture from those with HD due to end-stage liver disease, i.e., progressive bile duct obliteration and parenchymal fibrosis. However, T2-MRCP findings in the five PFS patients with histologically confirmed HD of advanced PSC showed the classic pruned tree appearance of the intrahepatic bile ducts. Therefore, the combined features of T1-MRC, T2-MRCP, and all other MR images vastly improved the distinction between FS and/or HD [15], as has previously been described [15, 16].

In equivocal clinical cases, a confident diagnosis on T1-MRC can sway the decision whether or not to perform ERCP. In this sense, T1-MRC can help triage patients for ERCP. In routine clinical practice, the management of PFS, based on a multiparametric MR exam, should include interdisciplinary discussion between experienced radiologists and gastroenterologists to reduce the burden of unnecessary ERCPs, as recommended by the IPSCSG [8].

Moreover, the superiority of multiparametric GA-MRI, providing simultaneous anatomic plus global/segmental functional hepatobiliary information, using functional liver imaging score (FLIS), can show PSC’s progression to end-stage cirrhosis with portal hypertension [11, 18‐20].

Importantly, in our study, we did not detect any CCA in the PFS diagnosis. The two (1.55%) confirmed cases of CCA in our cohort occurred in the NFS group. Our results are similar to those of Villard et al where very few of their cohort, i.e., 2%, had CCA and none in a significant stricture [44]. The real burden of CCA in a DS or FS should be further evaluated in a prospective multicenter study.

We acknowledge the inherent limitations of our retrospective study. First, by considering ERCP not only a secondary event, but also as the gold standard, our study had an indication bias. However, our main goal was to see if T1-MRC could yield an accurate and confident imaging diagnosis with better agreement among radiologists, i.e., a less subjective method than measurements (e.g., DS) or approximations (e.g., HGS). Therefore, we had to use ERCP to confirm our diagnoses which then led to either dilatation and/or stenting, if FS was diagnosed. This may have caused an overestimation of PFS events. Therefore, we advise caution in interpreting our prognostic results. Secondly, we must acknowledge that we used the term DS to interpret T2-MRCP images, despite the MR Working Group of the IPSCSG’s recent recommendation that the term DS should not be used with MRCP [8]. Thirdly, regarding our primary endpoints, we had almost 15% of our cohort who either underwent OLT or had liver-related mortality. We believe this relatively high percent of primary adverse events within a short follow-up time is because our tertiary care patients were severely ill compared to the average PSC patient. This is a well-known phenomenon in tertiary centers [45]. Fourthly, our retrospective study design, by skewing our inclusion cohort, caused rather short mean and median follow-up times (43.5 months and 30.7 months, respectively) as compared to our long observation period. Fifthly, we gave the results based on the analysis of the most experienced reader (almost perfect) according to the IPSCSG criteria [8] which may have led to an overestimation; however, the difference between experts and residents is negligible, as the median values for all readers show. Additionally, the median values also correlated with the events. Evenso, we advise caution in interpreting our prognostic results. Sixthly, all but three patients in the impaired excretion group underwent ERCP, due to suspected end-stage PSC, which was histologically confirmed at OLT or post mortem. Not all patients underwent ERCP because of its invasive nature. Our physicians only perform ERCP when clearly clinically indicated to avoid complications. However, no functional stricture (NFS) patient had subsequent loss of liver function or any event 6 to 12 months after MRI. Finally, liver stiffness for PSC severity prediction was measured with MR elastography in only one-third of our cohort. Our results showed no statistically significant difference in liver stiffness between NFS and PFS patients nor between events and no-events, likely due to the relatively small group [33]. Even so, MR elastography cannot diagnose FS.

In conclusion, T1-MRC allows a more confident and reproducible diagnosis and localization of stricture than T2-MRCP. PFS correlated better than DS or HGS with lab values, clinical scores, and outcomes.