The Prognostic Utility of Quantitative Magnetic Resonance Cholangiopancreatography in Patients With PSC: A Systematic Review With Structured Evidence Synthesis

Abstract

Background:

Magnetic resonance cholangiopancreatography (MRCP) is used in the diagnosis and management of primary sclerosing cholangitis (PSC). However, MRCP is subjective, and guidelines state that there is insufficient evidence to recommend MRCP as a prognostic tool. EASL-PSC guidelines have identified quantitative MRCP (MRCP+) as having promising utility for risk assessment and prediction of clinical outcomes. We conducted a systematic review to determine the clinical utility of MRCP+ metrics for the management of patients with PSC.

Methods:

We systematically searched PubMed and MEDLINE for cohort studies reporting on MRCP+ metrics associated with risk prediction, clinical outcomes, and disease progression published between January 2019 and May 2024. Studies reporting on adults with diagnosed PSC with paired MRCP were selected. There were no limitations on the type of study (retrospective/prospective). Studies reporting on only qualitative assessment of MRCP were excluded.

Results:

Six manuscripts with 512 subjects met the study criteria. For risk prediction and disease progression, 4 articles reported 21 unique MRCP+ metrics, which were significantly associated with the Amsterdam-Oxford model and the MAYO risk score. MRCP+ metrics had AUC ranging 0.65 to 0.87 and hazard ratios ranging 0.96 to 17.79 for the prediction of adverse outcomes, including liver transplantation, death, hepatic decompensation, biliary complications, and cholangiocarcinoma. Associations between biochemical markers of liver function (alkaline phosphatase, bilirubin, aspartate aminotransferase, gamma glutamyl transferase, and albumin) ranged -0.42≤R≤0.48.

Conclusion:

MRCP+ metrics have clinical utility to support patient management alongside addressing key gaps, including standardising MRCP assessment, early detection of disease, and quantification of risk.

Primary sclerosing cholangitis (PSC), a chronic progressive cholestatic liver disease characterized by multifocal biliary stricturing,¹ has no approved pharmacotherapies. Accurately characterizing the disease course is of critical importance for developing new, effective, and affordable treatments and for enabling timely intervention.² Currently, liver transplantation is the only lifesaving intervention, with recurrence after liver transplantation in 20% to 35%.² Recently, PSC has been declared one of the biggest unmet needs in hepatology, making up almost 6% of all chronic liver diseases (without hepatocellular carcinoma) requiring liver transplant.³ Identifying those at highest risk of poor outcomes is an integral part of the clinical assessment. Furthermore, opportunities for appropriate monitoring and intervention are limited due to the rarity of symptoms in early-stage PSC.²

For monitoring of disease progression, routine blood markers [alkaline phosphatase (ALP), bilirubin, albumin, platelets, and transaminases] are recommended.² Considering prognosis, ALP has shown the most clinical utility among the blood markers, with a cutoff of <1.5 upper limit of normal (ULN) associated with a lower risk of adverse outcomes.² However, as ALP can be affected by other factors, including treatment,² clinical guidelines do not recommend its use as a standalone marker for prognostic use. PSC with disease progression develop advanced fibrosis and cirrhosis with portal hypertension requiring liver transplantation, parenchymal liver fibrosis assessment using both vibration-controlled transient elastography (VCTE) and blood-based biomarker of fibrosis, enhanced liver fibrosis (ELF) have shown promise to support patient management.^2,4 Nevertheless, both VCTE and ELF are not recommended for use in isolation to manage patients and thus should be used in conjunction with other biochemical and radiological features.⁴

Magnetic resonance cholangiopancreatography (MRCP) is commonly used for the diagnosis of PSC, allowing anatomic characterization of the biliary tree, with many centers adopting annual MRCPs for monitoring of their patients for disease progression.² Clinical guidelines recommend the use of ultrasound in large duct PSC to monitor for gallbladder polyps and in liver cirrhosis for surveillance for liver lesions. MRCP has been shown to be superior to ultrasound for early detection of cholangiocarcinoma (CCA), although its role in screening for asymptomatic PSC-CAA is more controversial.⁴ Despite being the recommended diagnostic tool for PSC and its widespread use, there is notable variation in the assessment of MRCP data sets across radiologists and centers, with intercenter reproducibility reported as a limiting factor.² To address the anatomic depiction issues, clinical guidelines now call for the use of high-quality MRCP for the diagnosis and management of PSC. Unfortunately, the limitations pertaining to subjective assessments and unstandardized acquisition protocols persists.² These limitations and insufficient evidence prevent MRCP from being recommended as a prognostic tool. Predictive models, including the ANALI score, MAYO risk score, and Amsterdam-Oxford model (AOM), have been created to address this gap. The MAYO risk score, which is the most frequently used, provides short-term prediction of transplant-free survival (<4 y), while the AOM provides longer-term prediction.² However, despite the prognostic potential of these models, clinical guidelines do not currently recommend their use as more evidence (including head-to-head comparisons) is still required to reliably estimate prognosis and warrant their recommendation.^2,4

Clinical guidelines highlight key unmet needs in the management of PSC.² First, as PSC natural progression varies, and risk is difficult to predict, there is a need for the development of quantitative tools and techniques to monitor patients.² Second, due to the subjectivity associated with current MRCP assessment hampering intercenter reproducibility, and the use of unstandardized protocols, there is a need for objective tools which can help to standardize image acquisition and reporting.² This is particularly important as it will lead to the establishment of surrogate markers that reliably estimate prognosis and can be used to assess disease progression alongside predicting risk and clinical outcomes.² Quantitative MRCP (MRCP+) is a commercial software tool that models the biliary tree and provides robust quantitative markers that can change dynamically as the disease progresses. To this effect, the recent EASL-PSC guidelines have identified MRCP+ as having promising utility for risk assessment and prediction of clinical outcomes.² Subsequent recent evidence appears to show utility for metrics derived from this tool in the management of PSC.⁵

Our aim was to perform a systematic literature review (SLR) and evidence summary to assess the quantitative MRCP evidence to understand the clinical utility of MRCP+metrics for (1) risk prediction (prognostic capability), (2) prediction of clinical outcomes, (3) assessing disease progression.

METHODS

This systematic review was conducted and reported in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement. This study was registered on PROSPERO.

Criteria for Considering Studies for the SLR

Patient Population

The patient population included in this SLR covered those included in studies reporting data on adults with diagnosed PSC with paired MRCP. Studies reporting both prospectively and retrospectively collected data were eligible for inclusion.

Index Tests

The index test (intervention) of main interest was quantitative MRCP performed with the quantitative biliary imaging tool MRCP+ (Perspectum, Oxford, UK). Results from comparison with serum-based biomarkers (ALT, AST, ALP, bilirubin, etc.), and risk scores (MAYO, AOM, ANALI, etc.) were also computed, where data were reported.

Reference Standard

Only studies reporting quantitative MRCP (MRCP+) metrics derived from noncontrast 3D MRCP in a PSC population were eligible for inclusion. There were no limitations put on the type or number of quantitative metrics; however, studies reporting on only qualitative assessment of MRCP images were not eligible for inclusion.

Target Conditions (outcomes)

There were multiple target conditions (outcomes) for this study, including risk stratification, prediction, association with clinical outcomes, association with disease severity, and quantification of change over time (Supplementary Table 1, Supplemental Digital Content 1, http://links.lww.com/JCAT/A59). A wide range of clinical outcomes were accepted (including liver/PSC-related death, all-cause mortality, cholangiocarcinoma, decompensated cirrhosis, liver transplantation, cirrhosis, ascites, jaundice, end-stage liver disease, pruritis). Risk was defined using established risk scores and/or biochemical markers, with no limitations on disease severity; thus, all clinician-directed descriptions from the studies were included. To control for relevant populations, those with concomitant disease with atypical natural history compared with PSC (eg, Wilson disease) were excluded, while those with similar disease progression (such as autoimmune sclerosing cholangitis) were eligible for inclusion (Supplementary Table 2, Supplemental Digital Content 2, http://links.lww.com/JCAT/A60).

Study Design

All study designs were considered if they were reporting on patients with PSC undergoing noncontrast MRCP, which were subsequently analyzed with MRCP+. We restricted inclusion to English-language articles. We included studies reporting on a wide range of clinical outcomes, patient risk stratification (using MAYO, ANALI, AOM scores), and natural progression (change over time) (Supplementary Table 1, Supplemental Digital Content 1, http://links.lww.com/JCAT/A59).

To fulfill the aims of the review, we used an exhaustive list of MESH terms (Supplementary Table 2, Supplemental Digital Content 2, http://links.lww.com/JCAT/A60) including “primary sclerosing cholangitis,” “quantitative magnetic resonance cholangiopancreatography,” “biliary,” “prediction,” “change over time,” and “increased risk,” to search electronic databases for abstracts published between January 2019 and May 2024. Search results were combined into a Mendeley database to facilitate reference management. The last search was performed on May 21, 2024.

Risk Scores and Their Definitions

To investigate the relationship between MRCP+metrics and risk prediction, we evaluated MRCP+ metrics against 3 validated prognostic models used in PSC: the MAYO risk score, the Amsterdam-Oxford Model (AOM), and the ANALI scores.

The MAYO risk score is a model that is used to estimate short-term (∼4-year) transplant-free survival using age, total bilirubin, albumin, AST, and history of variceal bleeding. Higher scores indicate greater mortality risk.⁶

The AOM predicts medium-term to long-term transplant-free survival using routinely available variables. Core predictors include age, bilirubin, albumin, ALP, and platelet count. Higher scores (>2.0) indicate a high risk of liver transplantation (LT) or death for patients.⁷

The ANALI scores (both with and without Gadolinium) are MRCP-based radiologic scores that combine biliary ductal dilatation severity with signs of chronic liver disease (eg, portal hypertension), to stratify risk of radiologic progression and clinical outcomes.² Similar to both the MAYO and AOM scores, higher ANALI scores are strongly associated with adverse outcomes.⁸

Data Analysis

Data Extraction and Risk-of-Bias Assessment

Relevant information was extracted from the included studies by 2 reviewers, and conflicts were resolved by discussion. Data extraction included the recording of study design, participant characteristics, MRCP details, MRCP+ metrics, outcomes collected, and outcome results.

Quality and Risk-of-Bias Assessment

The QUIPS (QUality In Prognosis Studies) tool⁶ was used to assess study risk-of-bias. This tool was modified to suit the requirements of this review based on the Newcastle-Ottawa scale criteria.^7,8 As the modified tool has 6 domains (study population, attrition, procedure, confounders, outcome measurement, and outcome reporting), for each study, the adequacy of reporting was assessed by 2 people on 3 levels of risk bias—“yes,” “partial,” or “no.”

Following assessment and conflict resolution by the 2 reviewers, studies were assigned to 3 overall categories of risk-of-bias. The GRADE (Grading of Recommendations, Assessment, Development and Evaluations) approach⁹ was used to rate evidence quality. Low risk-of-bias was used to describe studies for which all domains had no risk, moderate risk-of-bias was used to describe studies for which either one or more domains were scored as “partial” or one domain was scored as “yes,” while high risk-of-bias was used to describes studies for which more than one domain was scored as “yes” for risk-of-bias between the 2 reviewers.

Two coauthors are employees of Perspectum Ltd (M.P. and E.S.). To mitigate potential bias, in addition to using the abovementioned tests, screening, extraction, and risk-of-bias ratings were performed with full oversight from academic coauthors (R.V., F.A., E.L.C., D.S.P.). Any disagreements between the reviewers were discussed and resolved through consensus by referring to the full text. The data was under the control of an individual (R.V.) without potential bias.

To minimize reporting bias, we prespecified inclusion of all study-reported associations between MRCP+ metrics and prespecified outcomes/risk scores; we did not down-select metrics post hoc.

RESULTS

Study Selection, Cohort Characteristics, and Evidence Quality

Study Selection and Cohort Characteristics

Eight studies were identified in the search for articles. Following screening titles, abstracts, and full text, 6 articles (N=512 participants, 1237 patient follow-up years) met the inclusion criteria for the SLR and evidence summary (Fig. 1). Of the 2 excluded studies, one did not report quantitative metrics while the other did not include the study population.

FIGURE 1.

Preferred reporting items for systematic reviews and meta-analyses (PRISMA) flow diagram. Meta-analysis was not performed as part of this review.

All eligible studies included adult populations with 4 single-center retrospective studies^10–13 and 2 prospective single-center studies.^5,14 The studies covered are representative of the PSC patient population with a median reported age (at diagnosis or at MRCP): 40 y and 65% male (Table 1). The main concomitant diseases were inflammatory bowel disease (55%; 281/512), ulcerative colitis (38%; 197/512), and the Crohn disease (14%, 67/512). In the studies evaluated, 29% (149/512) were reported as taking ursodeoxycholic acid (UDCA) as part of their management (Table 1).

TABLE 1.

Demographic Table Showing Population Characteristics From the Studies Included in the Systematic Literature Review

						Type of PSC			Associated Disease			Treatment
References	Age at Diagnosis, Median (Range) (y)		Male, n (%)	Number of Patients	Intrahepatic Only, n (%)		Both Intrahepatic and Extrahepatic, n (%)	IBD, n (%)	Ulcerative Colitis, n (%)	Crohn Disease, n (%)	Indeterminate Colitis, n (%)	Ursodeoxycholic Acid (UDCA), n (%)
Cazzagon et al10	34 (21-43)		52 (67.5)	77	24 (31.2)		53 (68.8)	58 (75.3)	31 (53.4)	23 (39.7)	4 (6.9)
Cristoferi et al11	32 (23-47)		53 (60.9)	87	27 (31.2)		60 (69.0)	54 (62.1)	39 (72.2)	13 (24.1)	2 (3.7)
Ismail et al12	41 (30.5-60)*		42 (64.6)	65
Selvaraj et al14	44 (31-56)†		51 (67)	76	35 (46.1)		41 (53.9)	57 (75)	42 (73.7)	8 (14)	7 (12.3)	42 (55)
Trivedi et al5	39.6 (18-68.9)†		33 (60)	55	31 (56)		20 (36.4)					22 (40)
Vuppalanchi et al13	48 (32-63)		102 (67)	152	47 (31.2)		105 (69.1)	112 (74)	85 (76)	23 (23)	1 (0.9)	85 (57.1)
Total N			512	281	164		279	281	197	67	14	149

^*Patient age is reported either at the time of MRCP.

^†Time of inclusion in the study.

Quality of Evidence

Following use of the QUIPS tool (modified based on the Newcastle-Ottawa scale criteria), of the studies assessed, 3 were high-quality while 3 were deemed to be of moderate quality (Supplementary Table 3, Supplemental Digital Content 3, http://links.lww.com/JCAT/A61). Technical failure of MRCP+ was consistently reported and was found to be low (<5% for prospectively collected images and <20% for retrospectively collected images). Prediction of risk and disease progression were judged to have high-quality evidence due to the frequency of reporting and the removal of potential bias following inclusion of cohort characteristics in the reported analyses.

MRCP+ Metrics for Risk Prediction

Association With Risk Scores

In relation to risk prediction, although slightly different (in terms of prediction horizon), both the Amsterdam-Oxford model (AOM) and MAYO risk scores are predictive models for transplant-free survival in PSC patients. Of the included studies, 4 articles reported that 21 MRCP+ metrics (volume, duct width, stricture, and dilatation) were significantly associated with the AOM model and the MAYO risk score (Table 2). In relation to the prediction of disease progression, the ANALI score was used to assess the prognostic capacity of MRCP+ metrics to predict disease progression. Extracted data showed that 3 metrics (number of dilatations, sum of relative severity of dilatations, and tree dilation score) were significantly associated with the ANALI risk score both with and without Gadolinium (Table 2; Supplementary Table 4, Supplemental Digital Content 4, http://links.lww.com/JCAT/A62).

TABLE 2.

Correlations Between MRCP+Metrics, Risk Scores (Amsterdam-Oxford Model, MAYO Risk Score, and ANALI Risk Score) and Liver Stiffness

		Amsterdam-Oxford Model		Mayo		ANALI w/ Gadolinium		Liver Stiffness
References	Metric	Correlation	P	Correlation	P	Correlation	P	Correlation	P
	Volume Metrics
Selvaraj et al14	Biliary tree volume (mlm-1)*	R=0.39	<0.01
Cazzagon et al10	Biliary tree volume					R=0.27	0.03
Trivedi et al5	Total number of modeled ducts							R=0.32	<0.02
Cazzagon et al10	Median length of ducts (mm)							R=-0.27	<0.02
	Duct Width Metrics
Cazzagon et al10	Duct diameter median (mm)			R=0.34	<0.01	R=0.65	<0.01	R=0.51	<0.01
Cazzagon et al10	Ducts with a diameter between 1 and 3 mm (%)			R=-0.36	<0.01	R=-0.45	<0.01	R=-0.47	<0.01
Cazzagon et al10	Ducts with a diameter between 3 and 5 mm (%)			R=0.27	0.03	R=0.48	<0.01	R=0.39	<0.01
Cazzagon et al10	Ducts with a diameter between 5-7 mm (%)	R=0.37	<0.05					R=0.29	0.01
Cazzagon et al10	Ducts with a diameter between 7-9 mm (%)	R=0.46	0.01	R=0.43	<0.01			R=0.25	0.03
Ismail et al12	Ducts with a diameter >9 mm (%)							beta=1.67	<0.01
	Combined Metrics
Cazzagon et al10	Ducts with ≥1 stricture (n)					R=0.34	<0.01
Cazzagon et al10	Ducts with ≥1 dilatation (n)					R=0.34	<0.01
Cazzagon et al10	Duct with ≥1 stricture OR dilatation (n)					R=0.34	<0.01	R=0.24	<0.04
Trivedi et al5								R=0.44	<0.01
Cazzagon et al10	Duct with ≥1 stricture AND dilatation (n)					R=0.31	0.01
Cazzagon et al10	Duct with ≥1 stricture OR dilatation (%)			R=0.38	<0.01	R=0.43	<0.01	R=0.24	<0.04
Cazzagon et al10	Duct with ≥1 stricture AND dilatation (%)
Cazzagon et al10	Total length of strictures and dilatations					R=0.36	<0.01
Cazzagon et al10	Proportion of the biliary tree that is abnormal (%)			R=0.39	<0.01	R=0.43	<0.01	R=0.37	<0.01
	Stricture Metrics
Selvaraj et al14	Total number of strictures (m-1)*	R=0.29	0.01	R=0.25	0.03			R=0.26	<0.03
Trivedi et al5	Total number of strictures							R=0.54	<0.01
Cazzagon et al10						R=0.35	<0.01
Cazzagon et al10	Total stricture length					R=0.31	0.01
Cazzagon et al10	Maximum absolute severity of strictures
Cazzagon et al10	Sum of absolute severity of strictures					R=0.34	<0.01
Cazzagon et al10	Number of strictures with a length <3 mm					R=0.29	0.02	R=0.35	<0.01
Cazzagon et al10	Number of strictures with a length between 3 and 10 mm					R=0.33	<0.01
Ismail et al12				Beta=0.04	<0.01
Cazzagon et al10	Number of strictures with a length ≥10 mm					R=0.25	0.05
Cazzagon et al10	Stricture score mean	R=0.41	P=0.02
Cazzagon et al10	Stricture score sum					R=0.30	0.02
Cazzagon et al10	Mean relative severity of strictures					R=-0.28	0.03
Cazzagon et al10	Sum of the relative severity of strictures					R=0.34	<0.01
Selvaraj et al14	Sum of relative severity of strictures (m-1)*	R=0.28	0.02	R=0.26	0.03			R=0.27	0.02
Selvaraj et al14	Proportion of the tree with strictures (%)							R=0.24	0.04
Selvaraj et al14	Tree stricture score							R=0.24	0.04
Ismail et al12	Minimum stricture diameter (mm)	Beta=-0.59	<0.01	beta=-0.57	0.04
	Dilatation Metrics
Selvaraj et al14	Total number of dilatations (m-1)*	R=0.40	<0.01	R=0.37	<0.01			R=0.31	<0.01
Cazzagon et al10	Total number of dilatations					R=0.35	<0.01
Cazzagon et al10	Total dilatation length					R=0.35	<0.01
Cazzagon et al10	Maximum diameter of a dilatation					R=0.39	<0.01	R=0.23	<0.05
Cazzagon et al10	Mean absolute severity of dilatations	R=0.47	<0.01	R=0.37	<0.01			R=0.41	<0.01
Cazzagon et al10	Maximum absolute severity of dilatations			R=0.34	<0.01	R=0.36	<0.01	R=0.26	0.02
Cazzagon et al10	Sum of absolute severity of dilatations					R=0.39	<0.01
Cazzagon et al10	Mean relative severity of dilatations	R=0.44	0.01	R=0.39	<0.01			R=0.34	<0.01
Cazzagon et al10	Maximum relative severity of dilatations			R=0.39	<0.01	R=0.30	0.02	R=0.25	0.03
Cazzagon et al10	Sum of the relative severity of dilatations					R=0.38	<0.01
Selvaraj et al14	Sum of relative severity of dilatations (m-1)*	R=0.45	<0.01	R=0.38	<0.01			R=0.36	<0.01
Selvaraj et al14	Proportion of the tree dilated (%)	R=0.47	<0.01	R=0.26	0.02
Selvaraj et al14	Tree dilatation score	R=0.46	<0.01	R=0.27	0.02
Cazzagon et al10	Dilatation score max							R=0.25	0.03
Cazzagon et al10	Dilatation score mean							R=0.23	<0.05
Cazzagon et al10	Dilatation score sum					R=0.39	<0.01

Metrics are grouped by construct (volume, duct width, stricture, dilatation, combined/composite) and to avoid selective reporting, all associations provided by the source manuscripts are shown.

^*Metrics reported as being normalized to the total length of the measured biliary tree are indicated with an asterisk.

In addition to the risk scores and biochemical markers used to assess disease progression, liver stiffness measurement (LSM) has been used to assess disease progression, as it can indicate the presence of liver fibrosis. From the included articles, the total number of ducts, median duct diameter, number of strictures, and the presence of ducts with more than one stricture or dilatation, among others, were associated with higher LSM (Table 2). In addition, associations with biochemical markers of liver function (ALP, Bilirubin, AST, GGT, and albumin) were the most commonly reported and ranged -0.42≤R≤0.48 (Table 3).

TABLE 3.

Correlations Between MRCP+Metrics and Biochemical Markers of Liver Health

		Bilirubin		AST		ALP		GGT		Albumin		Platelets Count
References	Metric	Correlation	P	Correlation	P	Correlation	P	Correlation	P	Correlation	P	Correlation	P
	Volume Metrics
Selvaraj et al14	Biliary tree volume (mlm-1)*											R=0.23	0.05
	Duct Width Metrics
Cazzagon et al10	Duct length mean (mm)			R=-0.27	<0.02
Cazzagon et al10	Duct diameter median (mm)			R=-0.42	<0.01	R=0.44	<0.01	R=0.40	<0.01	R=-0.27	<0.03
Cazzagon et al10	Ducts with a diameter between 1 and 3 mm (%)			R=-0.42	<0.01	R=-0.38	<0.01	R=-0.35	<0.01	R=0.27	0.02
Cazzagon et al10	Ducts with a diameter between 3 and 5 mm (%)			R=0.34	<0.01	R=0.32	<0.01	R=0.32	<0.01
Cazzagon et al10	Ducts with a diameter between 7 and 9 mm (%)	R=0.28	0.01	R=0.29	0.01	R=0.35	<0.01			R=- 0.34	<0.01
Cazzagon et al10	Ducts with a diameter > 9 mm (%)	R=0.31	<0.01	R=0.26	<0.03			R=0.24	0.04
	Combined Metrics
Cazzagon et al10	Duct with ≥1 stricture OR dilatation (%)	R=0.30	<0.01	R=0.35	<0.01	R=0.48	<0.01	R=0.40	<0.01
Cazzagon et al10	Duct with ≥1 stricture AND dilatation (%)					R=0.31	<0.01	R=0.26	0.02
Cazzagon et al10	Proportion of the biliary tree that is abnormal (%)	R=0.25	<0.03	R=0.32	<0.01	R=0.44	<0.01	R=0.41	<0.01
	Stricture Metrics
Selvaraj et al14	Total number of strictures (m-1)*			R=0.23	<0.05	R=0.25	0.03
Trivedi et al5	Total number of strictures					R=0.35	<0.01
Cazzagon et al10	Number of strictures with a length ≥10 mm											R=0.25	0.03
Selvaraj et al14	Sum of relative severity of strictures (m-1)*			R=0.23	<0.05	R=0.26	<0.03
	Dilatation Metrics
Selvaraj et al14	Total number of dilatations (m-1)*			R=0.23	<0.05	R=0.37	<0.01	R=0.30	<0.01	R=-0.29	0.01
Cazzagon et al10	Maximum diameter of a dilatation					R=0.31	<0.01	R=0.29	0.01
Cazzagon et al10	Mean absolute severity of dilatations			R=0.33	<0.01	R=0.31	<0.01	R=0.28	0.01	R=-0.29	0.01
Cazzagon et al10	Maximum absolute severity of dilatations			R=0.25	0.03	R=0.36	<0.01	R=0.32	<0.01
Cazzagon et al10	Sum of absolute severity of dilatations							R=0.25	0.03
Cazzagon et al10	Mean relative severity of dilatations			R=0.39	<0.01	R=0.36	<0.01	R=0.37	<0.01
Cazzagon et al10	Maximum relative severity of dilatations			R=0.31	<0.01	R=0.36	<0.01	R=0.38	<0.01
Cazzagon et al10	Sum of the relative severity of dilatations							R=0.25	0.03
Selvaraj et al14	Sum of relative severity of dilatations (m-1)*			R=0.24	0.03	R=0.38	<0.01	R=0.32	<0.01	R=-0.29	0.01
Selvaraj et al14	Proportion of the tree dilated (%)					R=0.26	0.02			R=-0.29	0.01
Selvaraj et al14	Tree dilatation score					R=0.27	0.02			R=-0.26	<0.03
Cazzagon et al10	Dilatation score sum							R=0.25	0.03

^*Metrics reported as being normalized to the total length of the measured biliary tree are indicated with an asterisk.

Table 2 summarizes all associations reported by the original studies to minimize selection bias; however, it is possible that metrics representing related constructs may exhibit within-family collinearity.

Prediction of Clinical Outcomes and Disease Progression

Pooled descriptive analysis of the reported findings showed that nine MRCP+ markers were repeatedly reported as being associated with clinical outcomes (Fig. 2, Table 4). Among the biliary tree volume, stricture, dilatation, and individual duct metrics were the most reported in the prediction of clinical outcomes (Table 4). Overall, MRCP+ metrics had AUC ranging 0.65 to 0.87 (CI: 0.36-0.96) and hazard ratios ranging 0.96 to 17.79 for the prediction of adverse outcomes ranging in severity from biliary complications to CCA and death (Fig. 2, Table 4).

FIGURE 2.

Quantitative MRCP (MRCP+) metrics reported as being associated with adverse outcomes, including liver transplantation, death, hepatic decompensation, biliary complications, and cholangiocarcinoma (CCA).

TABLE 4.

Hazards Ratio and Area Under the Receiver Operating Characteristic Curve (AUC) for MRCP+ Metrics

References	Metric	Hazard Ratio (95% CI)	AUC	95% CI
Cazzagon et al10	Duct Width Metrics
	Duct diameter median (mm)	17.79 (2.37-133.80)	0.79	0.68-0.89
	Ducts with a diameter between 1 and 3 mm (%)	2.53 (1.09-5.86)
	Ducts with a diameter between 3 and 5 mm (%) #1	5.23 (1.03-13.42)	0.78	0.68-0.89
	Combined Metrics
	Duct with ≥1 stricture OR dilatation (%) #2	3.14 (1.07-5.86)	0.67	0.54-0.79
	Proportion of the biliary tree that is of abnormal length (%)	3.39 (1.33-8.65)	0.68	0.56-0.81
	Dilatation Metrics
	Maximum diameter of a dilatation	5.23 (2.03-13.43)	0.65	0.53-0.78
	Maximum absolute severity of dilatations #3	3.19 (1.23-8.23)	0.65	0.52-0.77
	Maximum relative severity of dilatations	4.7 (1.74-12.70)	0.69	0.36-0.81
Cristoferi et al11	Volume Metrics
	Biliary tree volume (ml) #4	1.05 (1.03-1.08)
	Total number of modeled ducts	1.06 (1.03-1.08)
	Combined Metrics
	Duct with ≥1 stricture OR dilatation (n) #2	1.03 (1.01-1.04)
	Duct with ≥1 stricture AND dilatation (n)	1.1 (1.05-1.14)
	Total length of strictures and dilatations	1.03 (1.02-1.05)
	Stricture Metrics
	Total number of strictures #5	1.05 (1.02-1.08)
	Total stricture length	1.08 (1.04-1.12)
	Sum of absolute severity of strictures	1.47 (1.26-1.71)
	Stricture score sum #6	1.22 (1.11-1.34)
	Sum of the relative severity of stricture #7	1.01 (1.01-1.02)
	Dilatation Metrics
	Total number of dilatations #8	1.02 (1.01-1.04)
	Total dilatation length	1.05 (1.03-1.07)
	Sum of absolute severity of dilatations	1.14 (1.08-1.21)
	Sum of the relative severity of dilatations #9	1.04 (1.02-1.05)
	Dilatation score sum	1.07 (1.04-1.10)
Selvaraj et al14*	Volume Metrics
	Biliary tree volume (mlm-1) #4		0.72	0.59-0.85
	Sum of the lengths of all ducts (mm)		0.85	0.77-0.94
	Gallbladder volume		0.85	0.77-0.94
	Stricture Metrics
	Total number of strictures (m-1) #5		0.76	0.65-0.88
	Sum of relative severity of strictures (m-1) #7		0.79	0.68-0.79
	Proportion of the tree with strictures (%)		0.7	0.55-0.84
	Tree stricture score		0.71	0.58-0.85
	Dilatation Metrics
	Total number of dilatations (m-1) #8		0.86	0.78-0.96
	Sum of relative severity of dilatations (m-1) #9		0.87	0.77-0.96
	Proportion of the tree dilated (%)		0.85	0.73-0.95
Vuppalanchi et al14	Duct Width Metrics
	Ducts with a diameter between 3-5 mm (%) #1	1.06 (1.03-1.10)
	Stricture Metrics
	Stricture score sum #6	0.96 (0.93-0.98)
	Dilatation Metrics
	Maximum absolute severity of dilatations #3	1.23 (1.01-1.50)

Metrics reported as being normalized to the total length of the measured biliary tree are indicated with an asterisk (*), while metrics reported in more than one study are marked with an octothorp (#).

Two of the included articles reported head-to-head comparison of risk scores (ANALI, MAYO) and composite scores including MRCP+ metrics (Table 5). In these articles, for the prediction of clinical outcomes, MRCP+ metrics combined with other clinical markers (number of strictures in the qMRCP-PSC score and ducts with a diameter between 3 and 5 mm in the M+BA score) were better independent predictors of disease severity when compared with both the ANALI and MAYO risk score (Table 5).

TABLE 5.

Performance of Composite Scores Including MRCP+ Metrics With Existing Risk Scores

References	Metric	C-Statistic	Hazard Ratio (95% CI)	P-value
Cristoferi et al11	qMRCP-PSC score Number of strictures Spleen length Age at MRCP Height Time from diagnosis to MRCP	0.78		<0.001
	ANALI w/o gadolinium >2	0.64	3.32 (1.53-7.19)	0.002
Vuppalanchi et al13	M+BA risk score: Ducts with a diameter between 3 and 5 mm Total bilirubin AST		5.76 (1.50-22.07)	0.01
	Mayo risk score		3.13 (0.67-14.62)	0.15

Sensitivity Analysis

All studies were single-center retrospective studies, thus potentially increasing the overall risk-of-bias. Nevertheless, the performance of metrics in each individual study was very similar, with minimal nonsignificant differences in performance and confidence intervals in the studies included. Due to the differences in outcomes collected, analyzed, and subsequently underreporting of all available MRCPs metrics, selection bias (of MRCP+ metrics) was considered the main source of potential bias.

DISCUSSION

MRCP has excellent accuracy for the diagnosis of PSC; however, it is not recommended in clinical guidelines for prognostication and monitoring due to subjective assessment. In this systematic review evaluating the clinical utility of quantitative MRCP in 3 clinical categories—prediction of risk, clinical outcomes, and disease progression—there were 3 main findings. First, for risk prediction, MRCP+metrics related to biliary volume, duct dimensions, strictures, and dilatations were related to established risk scores, including the widely used MAYO and AOM models. Second, 9 MRCP+ metrics were most frequently reported for predicting clinical outcomes, including hepatic decompensation, liver transplantation, biliary complications, CCA, and death, with HR up to 17.79. Third, in relation to disease progression, 3 MRCP+dilatation metrics were significantly associated with the ANALI risk score both with and without Gadolinium.

In this SLR, we reviewed the largest sample of its kind with quantitative MRCP assessment in a rare disease. The study included a diverse set of groups with data collected from the UK, Europe, and Northern America (Canada and the United States), with 55% with inflammatory bowel disease, 38% with ulcerative colitis, and 14% with the Crohn disease. Our findings are therefore relevant for patients typically managed in secondary care in these territories and may be applied in the development of new strategies or in the consolidation of existing practices in evaluating patients for referral to secondary care. Despite the rarity of the disease, all included studies reported on a similar cohort size ranging 55 to 152 participants and covering 512 patients over 1237 patient follow-up years. Overall, MRCP+ metrics showed good associations with established risk scores. With respect to short-term transplant-free survival, MRCP+ metrics describing the biliary anatomy (volume, duct width, stricture, and dilatation) were significantly associated with the MAYO risk score. Similarly, regarding the association with longer-term transplant-free survival, MRCP+metrics, primarily characterizing strictures and duct dilatation of the biliary tree, were reported as being significant.

Early disease detection and avoidance of downstream adverse events are of paramount importance in the management of patients, forming one of the main pillars of precision medicine. Therefore, clinical outcomes form the basis for the assessment of non-invasive tools (NITs), particularly regarding prognostic ability. Our findings showed that MRCP+metrics have good prognostic ability to predict clinical outcomes (including need for liver transplant, development of cancer, and death) with HR ranging from 0.96 to 17.79. The combination of NITs has been shown to increase sensitivity and specificity in a wide range of chronic liver diseases and has led to the development of markers such as FIB-4, FAST,¹⁵ and MAST.¹⁶ Two of the studies included in the SLR evaluated the combination of MRCP+ and blood metrics (M+BA risk score, qMRCP-PSC score) and found that the combination of imaging and blood metrics enhanced the prediction of clinical outcomes when compared with existing risk scores (MAYO, ANALI).

Quantification of disease progression is an area of great importance for PSC. As there are no approved therapies for this indication, objective assessment of disease progression is a key area where evidence is lacking. Traditional biomarkers typically used in hepatology are often static and do not change meaningfully with disease progression or therapeutic intervention. Examples include genetic variants such as PNPLA3 or autoantibodies (eg, AMA, ANA, SMA), and baseline histologic fibrosis stage. These markers inform disease susceptibility or provide diagnostic confirmation but lack sensitivity to longitudinal change or treatment response. In contrast, dynamic biomarkers (typically imaging-based) reflect underlying pathophysiological processes that evolve with disease activity. Their continuous and reproducible nature allows for serial assessment of disease progression or regression, making them valuable for both prognostication and monitoring. Therefore, clinical guidelines recommend clinical review, standard liver tests every 6 to 12 months, annual liver ultrasound and/or MRCP, as well as liver elastography or serum fibrosis tests (ELF) every 2 to 3 years. The general goal is early detection of disease complications. With respect to imaging, although MRCP is the current diagnostic reference standard, in its current form (qualitative assessment) it is not recommended as for liver disease progression monitoring. Thus, determining the prognosis of people with PSC remains a challenge. Prognosis is further complicated by the highly heterogeneous nature of PSC, non-specific symptoms (jaundice, pruritis, and upper quadrant abdominal pain), the unsuitability of MRCP in a proportion of the patient population (∼10%) with peripheral intrahepatic branches that are not visible to the naked eye on imaging due to tapering and duct distortion, and unpredictable risk of epatobiliary malignancy.

Diagnostic stratification can be a bottleneck, as a range of other chronic liver diseases associated with biliary stricturing and elevated serum markers of cholestasis, including AIH-PSC overlap, IgG4-RC, and other causes of secondary sclerosing cholangitis, can show similar findings on MRCP. MRCP+metrics have shown utility to stratify between closely related diseases, with metrics such as the number of strictures having good performance (AUC: ∼0.8^10,11,17) to distinguish between PSC and other liver diseases.

As PSC is a stricturing disease, and the longitudinal qualitative assessment of strictures is subjective, quantitative assessment can yield valuable information pertaining to the natural history of PSC. For instance, in one study, a duct diameter of 1 to 3 mm was reported as having a Hazard ratio of 2.53. In the reporting paper, the % of ducts 1 to 3 mm showed an inverse association with adverse outcomes. Biologically, this is consistent with PSC progression, where diffuse ductal remodeling (loss of normal 1 to 3 mm ducts alongside increases in strictures and/or dilatations) signals more advanced disease. Therefore, although in the arguably normal range, in the background of PSC, this metric can provide clinically useful information. Another paper was identified that reported longitudinal changes in MRCP+ metrics in PSC. The number of strictures was shown to increase by 1 per year, an increase reported as significant (P=0.02) over one year. In another study, the hazard ratio relating to the likelihood of a hard outcome increased significantly (P<0.001) by 1.06 per-stricture, with good discriminatory power (C-statistic: 0.72). The number of strictures had good power for predicting all clinical events (hazard ratio: 1.05, P<0.001, C-statistic: 0.78).^10,20 The development of CCA is an ever-present risk for patients with PSC, with large studies showing that CCA can account for ∼32% of PSC-related mortality. Various risk factors for the development of CCA have been noted in current clinical guidelines, including rapid clinical deterioration, elevated CA 19-9 or high-grade stricture upon imaging; however, there are currently no appropriate screening methods that are safe, have high sensitivity, and are accepted by patients. Preliminary evidence has shown that MRCP+measurement of biliary volume has utility to differentiate between benign and malignant CCA.¹³ In addition, changes in MRCP+ metrics have been shown to impact clinical decision-making in the identification of high-grade strictures.¹⁴ These streams of research should be explored, and results published as they will form the much-needed evidence base showcasing the clinical impact this tool can have on patient management. The cost implications of including MRCP+ in the management of patients should also be evaluated. These analyses should include the time spent to evaluate cases alongside the potential reduction in technical failure (including the proportion of re-scans avoided). Furthermore, cost-effectiveness analyses, including the cost of implementation, should be conducted to further understand the feasibility of clinical implementation.^18–22

There are several strengths and some limitations to this work. First, the studies included are from a wide range of centers and use both retrospective and prospective MRCPs and thus highlight the versatility of both the MRCP+tool and the metrics derived from it. In terms of limitations, this SLR included a small number of papers with inherent heterogeneity in study design. This is not a novel finding, however, as PSC is a rare disease. Nevertheless, the included sample of articles is comparable to the SLR and evidence synthesis, which showed that MRCP had diagnostic utility for diagnosing PSC²³ and the utility of imaging indices and features to predict PSC progression.²⁴ The summary results of this review had some statistical heterogeneity, which can have multiple sources, including differences in the test or study procedures, subject populations, study designs, or a combination of these factors. Although we report here findings from an SLR of published literature, a pooled independent patient data meta-analysis should be performed to enable reporting of a select group of metrics with prognostic utility in PSC. Furthermore, these investigations should explore the meaningful change in MRCP+metrics relating to natural history and disease progression, as these can support drug development. This is particularly so as only one paper reported disease progression and reported a change to the number of strictures over 1 year (P=0.02), with no changes in either liver stiffness measurement, ANALI score, or biochemical measures. Lastly, in the context of drug development, the combination of multiple modalities, as shown in early work²⁵ has the potential to capture the resultant changes in liver parenchyma following therapeutic intervention. Although only preliminary data exist to this effect, future studies should evaluate the effects of therapeutic intervention on both the biliary tree and liver parenchyma surrounding the ducts.

In conclusion, we performed a systematic literature review and structured evidence synthesis with the aim of assessing the quantitative MRCP evidence presented in the literature pertaining to disease progression as well as risk and clinical outcomes prediction. Our findings showed that MRCP+metrics have clinical utility to support patient management in all three domains, alongside addressing key gaps, including standardizing MRCP assessment, early detection of disease, and quantification of risk.

Supplementary Material

rct-50-365-s001.docx ^{(16.4KB, docx)}

rct-50-365-s002.docx ^{(16.2KB, docx)}

rct-50-365-s003.docx ^{(15.9KB, docx)}

rct-50-365-s004.docx ^{(18.5KB, docx)}