The role of serology, liver function tests and imaging in screening of primary sclerosing cholangitis: the HelPSCreen score

Abstract Objecives At present, no sensitive or specific screening test exists for primary sclerosing cholangitis (PSC). PSC screening is mainly based on elevated alkaline phosphatase (ALP) in patients with inflammatory bowel disease (IBD). We aimed to produce a screening score based on laboratory tests to predict the likelihood of PSC. Moreover, we evaluated the additional roles of liver histology and magnetic resonance cholangiopancreatography (MRCP) in the diagnosis of PSC. Materials and methods The data of 385 patients who came for their first endoscopic retrograde cholangiography (ERC) to confirm PSC diagnosis were retrieved from the PSC registry of the Helsinki University Hospital. Overall, 69 patients referred for ERC with suspected PSC, in whom PSC was excluded by ERC or liver biopsy and MRCP, served as controls. We included patients’ demographics and 13 laboratory test results in the analysis. Variables with significant odds ratios were selected for multivariate logistic regression, which was used to create a novel scoring system for PSC. The presence of IBD, serum perinuclear anti-neutrophil cytoplasmic antibodies, and ALP levels demonstrated the highest predictive value for PSC. A score was assigned for each statistically significant predictor. Results The optimal cut-off point for the score was ≥3, with an AUC of 0.83 (95%CI: 0.78–0.88). The addition of liver histology or MRCP findings to the score did not add a predictive value. Concusions In conclusion, we created a novel, simple scoring system to screen the probability of PSC. The HelPSCreen-score may help to assess the disease prevalence and to target further investigations in patients suspected of PSC.


Introduction
Primary sclerosing cholangitis (PSC) is a chronic cholestatic liver disease.It affects the intra-and extra-hepatic bile ducts, causing inflammation that leads to bile duct strictures.Approximately, 76% of patients with PSC have concomitant inflammatory bowel disease (IBD) [1,2] and the risk of malignancy, especially cholangiocarcinoma (CCA), is markedly elevated, with a lifetime risk of approximately 20% [3][4][5].
There are only a few population-based studies on the incidence and prevalence of PSC; most of these studies are small case series from tertiary centers.PSC have an incidence rate of 0.07-1.6 per 100,000 person-years and a prevalence rate of 0.22-32 per 100,000 inhabitants [6,7].
Patients with PSC that are symptomatic at the time of diagnosis are ≥50% [1] while CCA is diagnosed in 27-50% of patients at the time of diagnosis or within a year of diagnosis [8,9].This underlines the importance of early diagnosis of PSC [10].
The preferred diagnostic test for PSC is magnetic resonance cholangiopancreatography (MRCP) or endoscopic retrograde cholangiography (ERC) in some cases [11][12][13][14].The diagnostic criteria for PSC have recently been updated [15].MRCP has been demonstrated to have a high diagnostic accuracy -both sensitivity and specificity [16].However, contradictory results have been published with a direct comparison to ERC [17].MRCP has low sensitivity for detecting early intrahepatic changes and even advanced extrahepatic lesions [17].
The latest guidelines from Europe and the United States do not recommend a liver biopsy, except in cases where small-duct PSC or overlap syndrome is suspected [11][12][13][14].Histological features of PSC are vanishing bile ducts, cholangitis and concentric periductal fibrosis [18,19].However, less than 20% of PSC liver biopsy specimens have the typical histological features, a scenario that can also be observed in cases of secondary sclerosing cholangitis [15].No internationally validated histological scoring system specific for PSC exists; however, the adapted Nakanuma scoring system, and particularly, the fibrosis stage, has been shown to have a strong predictive value for late endpoints such as PSC-related death and liver transplantation, with a substantial interobserver agreement in an international cohort study [20][21][22].
Currently, no specific diagnostic test, screening test or test suggestive of PSC is available, unlike serum anti-mitochondrial antibodies (AMAs) for primary biliary cholangitis (PBC) [19] and serum smooth muscle antibodies (SMAs) for autoimmune hepatitis (AIH) [23].Results from laboratory tests revealed that plasma alkaline phosphatase (ALP) and plasma gamma-glutamyl transferase (yGT) levels are often elevated; however, they fluctuate and may sometimes be normal [2,24,25].Serum immunoglobulin G4 (IgG4) is elevated in few patients with PSC [26]; however, an elevation of ≥4 times the upper limit of normal and an IgG4:IgG1 ratio higher than 0.24 are diagnostic for IgG4-associated cholangitis [27].Serum atypical perinuclear antibody (pANCA) can be elevated in approximately 26-94% [28] of patients with PSC; however, it is not specific for PSC.Elevated serum antinuclear antibody (ANA) and SMA are also common findings, but as with pANCA, they are not specific [29,30].
We aimed to construct a simple screening score based on demographic variables and laboratory tests to predict the likelihood of PSC.This will aid the diagnosis of PSC at an earlier stage and enable optimal targeting of further examinations, like MRCP and ERC.Moreover, we evaluated the added value of imaging and liver histology in the diagnosis of PSC.

Patients
We collected the data of patients with PSC from the PSC registry of the Helsinki University Hospital (HUH).The PSC registry is a comprehensive registry of all patients with PSC diagnosed within the HUH.It includes patients from the HUH area, which comprises 27% of the Finnish population, and patients referred from other areas in Finland.Patients whose data were retrieved had their first diagnostic ERC to confirm the presence of PSC.The indications for ERC were suspicion of PSC on MRCP and/ or at liver biopsy, constantly elevated ALP levels in concurrence with IBD, or elevated ALP and yGT levels in patients with negative AMA (Supplementary material, Figure 4).Patients with suspected PSC who underwent ERC and/or MRCP and liver biopsy (excluding PSC) served as controls (non-PSC patients).The patients in the non-PSC group were followed for at least 3 years to exclude any patients with early PSC in this group.The results from diagnoses of non-PSC controls are presented in the Supplementary material, Table 4.
ANCA antibodies were measured by indirect immunofluorescence IIF.cANCA was defined by diffuse granular staining of the cytoplasm, whereas pANCA was characterized by staining of the perinuclear cytoplasm [31].
According to the PSC histoscore, liver histology was scored using Herovici and cytokeratin 7 (K7) stained core needle biopsy specimens [32].It is a histological scoring method based on the adapted Nakanuma classification that was developed to evaluate histological features that are most likely to predict disease outcomes in PSC.Two experienced liver pathologists (S.B. and N.S.) evaluated the biopsy specimens obtained at the time of diagnosis for the quantity of fibrosis, portal inflammation, portal edema, hepatitis activity, pericholangitis, cholangitis activity, bile duct loss, ductular reaction and chronic cholestasis.Bile duct loss, ductular reaction and chronic cholestasis were evaluated using K7-stained specimens [32].In our PSC registry with more than 1000 patients with disease documented by ERC with balloon occlusion technique [17], not a single case with small duct PSC was diagnosed.
In our laboratory, as routine diagnostics, we applied a PSC likelihood score for every liver biopsy specimen to indicate the overall disease likelihood of PSC (Supplementary material, Table 5) [32].If laboratory parameters were available for patients (PSC and non-PSC) ± 3 months after the biopsy date, the patients were included in this study [32].
MRCP findings before diagnosis were scored from the existing radiology report as negative, suggestive or compatible with PSC.

Statistics
Descriptive statistics are presented as means with standard deviation (SD), as medians with interquartile range (IQR), or as counts with percentages (%).As appropriate, group differences were evaluated using the unpaired Student's t-test, Mann-Whitney U-test, Chi-squared test or Fisher's exact test.The area under the curve (AUC), positive predictive value, likelihood ratio, sensitivity, specificity and odds ratio (OR) were calculated.All determinants with p < .05were collectively entered into a multivariate logistic regression model to evaluate those that were independently associated with the presence of PSC.
The score assigned to each statistically significant predictor regression (ln(OR)) coefficient was rounded to the nearest integer.The scores were then summed to create a HelPSCreen score for each participant.Measures of goodness of fit of each diagnostic model were assessed using the Hosmer-Lemeshow test.The probability (%) of a true PSC diagnosis was based on the number of estimated score points (HelPSCreen) using logistic models.The normality of variables was evaluated graphically using the Shapiro-Wilk W-test.The Stata 17.0, StataCorp LP (College Station, TX) software was used for the analysis.

Study inclusion and patient characteristics
We retrieved the data of 1107 patients with large duct PSC from the PSC registry.Patients with PSC and without PSC were 982 and 125, respectively.Patients with PSC-AIH overlap syndrome were excluded from this study.The other reasons for exclusion are shown in Figure 1.After exclusion, 385 patients with PSC and 69 non-PSC patients were included in the analysis.The main characteristics of the patients and controls at the time of diagnosis are summarized in Table 1.Liver biopsy was available for 184 (48%) of patients with PSC and 35 (51%) of non-PSC patients.

Screening scores
Variables with significant univariate OR for differential diagnosis were chosen (Table 2).Multivariate logistic regression of the variables was performed, and a score was assigned to each statistically significant variable based on the regression coefficient (beta) rounded to the nearest integer (Supplementary material, Table 6).

Laboratory and IBD
A screening score based on IBD and laboratory results, the HelPSCreen score, included IBD, ALP and pANCA.The presence of IBD and pANCA gives two points each, and ALP > upper normal limit gives one point; yGT and ANA did not add a predictive value and were not included; hence, a maximum score of five points (Table 3).The optimal cut-off point for the score was ≥3, with an AUC of 0.83 (95%CI, 0.78-0.88)(Figure 2).The sensitivity, specificity, positive predictive value and OR for the score at this cut-off point were 0.71 (95%CI: 0.66-0.75),0.83 (95%CI: 0.72-0.91),0.96 (95%CI: 0.93-0.98)and 11.43 (95%CI: 5.96-21.91),respectively (Table 4).The distribution of HelPSCreen scores in PSC and non-PSC patients is shown in Figure 3(A).The probability of a true PSC diagnosis with a score of ≥3 was >90% (Figure 3(B)).

Magnetic resonance imaging (MRI) or MRCP, laboratory and IBD
MRCP reports at the time of diagnosis were available for a total of 412 (91%) patients: 347 (90%) from patients with PSC and 65 (94%) from non-PSC patients.Overall, 131 patients had negative MRCP findings, 152 were suggestive and 129 were compatible with PSC.Of the PSC patients, 95 (27%) had negative findings for PSC, whereas 29 (65%) of the non-PSC patients had MRCP findings suggestive or compatible with PSC.

Histology, laboratory and IBD
A total of 219 patients with liver biopsy specimens were available at the time of diagnosis: 184 patients with PSC and 35 non-PSC controls.
The AUC value (0.55 at a sensitivity of 0.1 and specificity of 1.0) was low for the liver biopsy specimens that have a likelihood score of 3 for PSC at the time of diagnosis.In cases where the likelihood score for PSC was 0, approximately 66.7% of the cases were still diagnosed with PSC using other modalities.
The HelPSCreen-H-score, a third score based on histological variables from the PSC histoscore (the stage of fibrosis, and quantity of portal inflammation) and logistic regression analysis of differential diagnostic ability, produced a better AUC value of 0.73 (95%CI: 0.66-0.80).In the multivariate logistic regression analysis, the OR for fibrosis stage and portal inflammation was 8.85 and 3.04, respectively.
A combination of histological variables and laboratory parameters (ALP, yGT and pANCA) did not improve the accuracy of the HelPSCreen score.

Discussion
Currently, there is no specific screening test for the diagnosis of PSC, and ERC, which is the gold standard, is invasive and costly.The recommended screening and diagnostic tool is MRCP; however, it has limitations in identifying early-stage intrahepatic and advanced extrahepatic bile duct diseases [16].In addition, there are reports of gadolinium accumulation in the brain in both children and adults with contrast-enhanced MRI, especially with repeated examinations [33] In this study, we evaluated a simple non-invasive screening test for PSC using systematically performed ERC and liver histology as the gold standard.In our center, ≥80% of patients were asymptomatic at the time of diagnosis [34].ALP and yGT were often elevated; however, they were not specific for PSC [14].ANCA is not disease-specific and can be detected in 50-85% of patients with ulcerative colitis [35,36] and 10-20% of patients with Crohn's disease.In PSC, ANCA positivity varies from 26 to 94% [24].PR3-ANCA has previously been shown to be a promising biomarker for PSC [37].Stinton et al. analyzed ANCA in 244 patients with PSC and 254 controls, including AIH, PBC, hepatitis C, hepatitis B and healthy controls.They found that PR3-ANCA was detected in 38.5% of patients with PSC compared to 10.6% in controls (p < .0001)and concluded that PR3-ANCA is detected in a significant proportion of patients with PSC compared to those with other liver diseases.PR3-ANCA was also associated with higher liver enzyme levels in patients with PSC [37].In this study, pANCA positivity was detected in 50% of PSC patients and 14% of non-PSC cases.In PSC patients with ulcerative colitis, 55% of the patients were positive for pANCA.Therefore, pANCA alone is not suitable for PSC screening.
Patients with PSC have a markedly increased risk of malignancy, particularly CCA and colorectal carcinoma.Therefore, it is important to diagnose patients with PSC at an early stage so that adequate surveillance can be implemented.
In our study, we included a panel of several laboratory tests and combined them with demographic data, imaging data and histology.We searched for the best combinations of these variables to predict the likelihood of having PSC.Based on these combinations, we produced the HelPSCreen score, a simple and noninvasive novel prognostic score to diagnose patients with PSC.Moreover, we analyzed the additional value of imaging (HelPSCreen-I-score) and histology (HelPSCreen-H-score).
The HelPSCreen and HelPSCreen-I scores both had an AUC >0.80, indicating good accuracy.The International Primary Sclerosing Cholangitis Study Group developed a consensus statement for PSC in 2021 [15], in which the diagnostic criteria are outlined.Large-duct PSC requires an MRCP or ERC compatible with PSC and one of the other criteria: IBD, serum markers for cholestasis, or histological features compatible The HelPSCreen-score-I supports these statements; however, the HelPSCreen-score does not, as it does not include MRCP.
The scores included readily available laboratory tests at low cost.IBD status is needed for the score; however, when suspecting PSC, most patients undergo a colonoscopy, as approximately 70% of patients with PSC have a concomitant IBD [8,38].Even in the absence of IBD, with a HelPSCreen-score of three points, the probability of PSC was more than 75%.Compared to present clinical practice evaluating the elevated ALP with or without concomitant IBD with MRCP, HelPSCreen offers a more accurate and cost-effective strategy for targeting further investigations.Patients with a positive score are referred for MRCP and ERC if indicated (Supplementary material, Figure 4).

Study limitations
The MRCP reports used were original reports from the hospitals where the patients were referred.Most of the reports were created by general radiologists, as all referral hospitals do not have radiologists specializing in hepatology.This can affect the yield of the MRCP, and the outcome might have been better for the HelPSCreen-I-score if the MRCP reports were evaluated by radiologists that specialized in hepatology.
In the early stages of PSC, the histological features may be patchy, and sampling error may be the reason why liver biopsy was not diagnostic [39].The same is true for follow-up.The PSC-histoscore, applied to score the liver specimens in this study, was able to show a predictive value for disease progression with endpoints of death, CCA or liver transplantation [32].As a diagnostic tool, it does not add a predictive value.One explanation for the poor histological yield is selection bias, as liver biopsy is often performed in patients whose diagnosis remains uncertain after MRI or MRCP and laboratory parameters.
In conclusion, despite its predictive value in the evaluation of disease progression, liver histology (HelPSCreen-H) did not improve the diagnostic accuracy in our cohort, not even in intrahepatic disease.
In HUH, ERC is performed to confirm the diagnosis of PSC, which provides a unique opportunity to study the correlation  between MRCP and ERC to build a screening score.As most other centers do not perform ERC to confirm the diagnosis, we did not have the opportunity to use a cohort from another center to validate our score.Further studies are needed to determine the performance of this score in other prospective cohorts.

Figure 2 .
Figure 2. auc and optimal cut-off value for the diagnosis of PSc (HelPScreen score) based on laboratory data and history of iBd.

Figure 3 .
Figure 3. (a) Bar chart showing the number of patients according to score (HelPScreen) points (total) and those with confirmed PSc diagnosis (grey bars) and non-PSc controls (white bars).(B) Probability (%) of PSc diagnosis based on the number of score points (HelPScreen).

Table 1 .
demography of non-PSc controls and patients with PSc at the time of diagnosis.

Table 2 .
Sensitivity, specificity, positive likelihood ratio and positive predictive values used for the screening and diagnosis of PSc.

Table 3 .
Multivariate logistic regression of clinical and laboratory variables with best discrimination values for differential diagnosis and their respective scores (HelPScreen).
a Score assigned to each statistically significant predictor regression coefficient rounded to the nearest integer.