Disease burden in primary sclerosing cholangitis in the Netherlands: A long‐term follow‐up study

Primary sclerosing cholangitis (PSC) is a progressive, cholestatic liver disease which greatly impacts the lives of individuals. Burden of disease due to shortened life expectancy and impaired quality of life is ill‐described. The aim of this study was to assess long‐term disease burden in a large population‐based registry with regard to survival, clinical course, quality adjusted life years (QALYs), medical consumption and work productivity loss.


| INTRODUC TI ON
Primary sclerosing cholangitis (PSC) is a chronic cholestatic liver disease characterized by multifocal biliary lesions. Inflammatory bowel disease (IBD), mainly ulcerative colitis (UC), coexists in 70% of PSC patients. 1 At diagnosis, 44% of patients had symptoms in a large Swedish cohort. 2 Pruritus, fatigue and right upper quadrant abdominal pain were most frequently reported. With disease progression, recurrent cholangitis and complications of cirrhosis can occur. 1 To date there is no curative therapy other than liver transplantation (LT), although patients might develop recurrent PSC. 1 PSC is associated with an increased risk of malignancies, mainly hepatobiliary malignancies such as cholangiocarcinoma and gallbladder cancer (GBC). 3,4 Risk of colorectal cancer (CRC) is fivefold higher in PSC-IBD patients compared to IBD only patients. 4 PSC is a severe disease with a dismal prognosis. The reported median time between diagnosis and LT or death varies from 14.5 years for tertiary centre cohorts to 21 years for population-based cohorts. 4,5 Studies regarding the natural history and disease management have provided some insight in the clinical events which PSC patients may encounter. However, benchmark studies containing quantitative data regarding quality of life (QoL), medical consumption and work productivity loss are lacking. Such data are essential for assessing the burden that these patients experience throughout their disease course and for development of new therapies in terms of healthcare evaluation studies. Accurate data on medical consumption are also indispensable for healthcare policy makers to allocate sufficient resources to the care of patients. Burden of disease is driven by shortened life expectancy and loss of QoL when alive. Hence, assessment of disease burden starts with accurate description of long-term survival.
The aim of this study was to assess long-term disease course with an emphasis on disease burden that patients experience during their extended patient journey in terms of survival, clinical events, QoL, medical consumption and work productivity loss.

| Study design and population
Patients in this observational dynamic registry were included from 2008 until 2020. Data were collected retrospectively from the date of diagnosis until 2008, and prospectively from that date onwards.
The study cohort consisted of the population-based EpiPSCPBC

| Patient involvement
Patients were involved in the development of both design and content of the study. Important study topics, frequency and extent of questionnaires and practical implementation were consulted with the patient organization. After every questionnaire, feedback of patients was evaluated and incorporated where possible. Progress of the study and interim results were biannually presented in digital news flashes and on patient conferences.

| Data collection
Data were collected retrospectively from the date of PSC diagnosis onwards (baseline), followed by annual medical chart review and/or medical letters from treating physicians. At baseline, demographics and clinical data regarding PSC diagnosis and coexisting IBD were collected. Alkaline phosphatase (ALP) >1.

| Clinical events
Incidence rates per 1000 patient years at risk were presented with 95%CI calculated using the Rothman/Greenland formula. 10 In case of zero events, the upper bound of the 95%CI was calculated using the Bayr's method. Incidence rates were calculated for large duct PSC, small duct PSC and patients with features of AIH (either large or small duct) separately. 95%CI of the difference between subgroups was calculated using the Bayr's approximation method and tested with the mid-P method using WinPEPI en OpenEpi software. 11,12

| Survival
The endpoint LT-free survival is a combined endpoint consisting of LT-or PSC-related mortality (death from end-stage liver disease, hepatobiliary malignancy or CRC). Cumulative incidence function (CIF) was calculated in R using the cmprsk package. 13  With regard to the endpoint LT-free survival, unrelated mortality was considered a competing risk. For LT or death from liver failureunrelated mortality and death by biliary malignancy was considered a competing risk. For biliary malignancy, all-cause mortality and LT were considered competing risks. (ii) the first year after LT is accompanied by substantial morbidity.

| Quality adjusted life years loss
The distribution of these various disease stages was assessed for each patient per 5 years after diagnosis and was derived from the entire registry. Health-related QoL in these distinct disease stages was assessed with the EQ-5D-5L health status questionnaire. Scoring profiles are assigned health utility weights based on an available Dutch health scoring algorithm 14  and Dutch EQ-5D-5L reference data, age and gender distribution of the cohort was taken into account. 14,15 Proportional shortfall (loss of QALYs relative to QALYs in the reference cohort) was calculated per 5 years after diagnosis.

| Medical consumption and costs
Medical consumption was measured with the iMTA Medical Consumption Questionnaire (IMCQ) 16

| Work productivity loss
Disease-related work productivity loss was measured repetitively with the IPCQ in the working population and pertained to the week prior to the data of the questionnaire. 20 The working population was defined as all patients who stopped working because of progressive PSC and/or IBD-the fully disabled-and patients at work, some of whom may be temporarily on sick leave due to PSC and/ or IBD complaints. Fully disabled patients were considered to have 100% productivity loss. Productivity loss while having a paid job was expressed as the percentage of normal working hours spent on sick leave. Data on work productivity loss are presented per 5 years after diagnosis before LT and <1 and >1 year after LT as mean percentage productivity loss of the working population and patients with paid work.

| Statistical analysis (other)
Pearson's chi-squared test was used for categorical data. Independent t-test, one-way ANOVA, Mann-Whitney U test or Kruskal-Wallis was used for continuous data, depending on the number of groups, the distribution of data and the variance (based on Levene's test).
Normality was assessed by normality tests and visual inspection of histograms and Q-Q plots. In case of median values of 0, because of highly skewed data, the median with interquartile range was not informative to report. Therefore, data were presented as mean (SD), but non-parametric tests were used for the analyses.

| Role of the funding source
The study was funded by ZonMW (grant number: 836041010). The funder had no role in study design, data collection, data analysis and interpretation, writing of the manuscript or decision to submit for publication. None of the authors was paid to write this article by a pharmaceutical company or other agency.

| Patients and response
A total of 1208 patients were included with a median follow-up of 11.2(5.9-17.6) years. Baseline characteristics of the entire cohort and the 359 patients participating in the PRO substudy are presented in Table 1. Baseline characteristics according to PSC phenotype are shown in Table S1. A sensitivity analysis comparing patients who participated in the PRO questionnaires versus those alive at the issuing of the first questionnaire but not participating is presented in Table S2. There was a small but significant decrease in the amount of patients with a coexisting IBD diagnosis, mostly UC, transplanted and an elevated ALP >1.3× ULN at diagnosis in the patients in the PRO substudy compared to the non-responders alive at the first questionnaire. Survival, clinical events and QALY loss were based on the entire cohort, medical consumption and costs, as well as work productivity loss was evaluated in patients returning PRO questionnaires. For various reasons, not all patients could be approached to receive PRO questionnaires ( Figure S1). Out of 511 invited patients, 359 (70%) agreed to receive questionnaires. Of those, 316 patients (91%) completed at least one questionnaire. During 3 years, 3525 questionnaires were sent and 67% were completed. To check for potential bias on reporting by phenotype, we determined whether completing more questionnaires was associated with certain factors.
Spearman correlation showed no significant bias, see Table S3.

| Clinical events
Incidence rate per 1000 patient years of decompensated cirrhosis was 16.9 in large duct as opposed to 3.4 in small duct PSC patients (p = 0.004) ( Table 2). Incidence rate of LT was 21.  (Table S4). During the prospective FU phase 2/76 (3%), small duct patients were documented to have progressed to large duct disease.   being alive after LT. In the first 30 years after diagnosis, PSC patients generated a total of 18.2 QALYs on average. This is 5.7 QALYs less than a Dutch age-and sex-matched reference cohort in the same period ( Figure 2B). Proportional shortfall of QALYs was 7% in the first 5 years after diagnosis and rose to 48% 26-30 years after diagnosis.

| Work productivity loss
The analysis regarding work productivity loss was based on the patients participating in the PRO substudy (N = 359). IPCQ was completed at least once by 324 patients, of which 68 (21%) had no paid work due to other causes, resulting in a working population of 256 patients. Patients with paid work had on average 34 contract hours per week. In Figure 5, mean productivity loss of the working population was 25%. Before LT, productivity loss increased from 18% to 34% (p < 0.001). Productivity loss among patients with paid work (working population without fully disabled patients) was stable over time. The grey area represents the survival curve of the entire cohort. It is important to notice that during follow-up, severely ill patients disappeared from the cohort due to death or LT.
TA B L E 2 Incidence rates (95% CI) per 1000 patient years at risk for clinical events in all patients and stratified for PSC type  F I G U R E 1 Cumulative incidence curves (95% CI intervals) for PSC-related mortality or LT (A), LT or death from liver failure (B) and biliary malignancy (C). The red line represents the cumulative incidence of the outcome respecting competing risks, the grey line represents the cumulative incidence of competing risk for reference.
Mean productivity loss of the working population in the first year after LT and >1 year after LT was 58% and 38% respectively (p = 0.085). Patients after LT with paid work had production loss of 43% in the first year decreasing to 9% >1 year post-LT (p < 0.001).

| DISCUSS ION
We here present the first longitudinal data on disease burden in PSC in the largest population-based registry to date. Disease burden is driven by both morbidity and mortality. LT-free survival in this study, taking into account competing risks, is slightly better than in most other population-based studies. 2,4,5,23 Survival until LT or all-cause mortality without taking competing risks into account was 18.0 years in our cohort, which is ~4 years longer than the same endpoint in the IPSCSG study which consisted mainly of tertiary care patients. 5 This is likely explained by selection bias. 2 Besides decreased life expectancy, patients experience substantial disease burden during the course of their illness. PSC is a slowly progressive, but ultimately severe disease, which is reflected by relatively low disease burden in the first years after diagnosis (9%) and  Hence, work productivity loss in PSC causes substantial economic burden.
The major strengths of this study lay in its large sample size with more than 1200 patients, its population-based nature, and the long median follow-up period of more than 11 years. With numbers more than doubled and a median FU 1.5 times longer we strengthen and recapitulate results from our earlier study confirming that prognosis is considerably better than that observed in predominantly tertiary care series. 2 coexisting IBD diagnosis in general, but not with regard to IBDphenotype, and an elevated ALP >1.3x ULN at diagnosis were slightly, albeit significantly, less represented among the patients in the PRO substudy. Furthermore, age at diagnosis was slightly higher in those taking part in the PRO study.
The generic EQ-5D-5L health-related QoL instrument was used to derive QALYs and assess the burden of disease of PSC patients from a societal perspective. Although generic, the EQ-5D-5L domain and index scores reflect the impact of symptoms as pain, itching and fatigue 34 which are recognized features of how patients experience having PSC in daily life. 35 The EQ-5D-5L scores undoubtfully would also reflect worries, depressed feelings and limited functioning in work and social activities to a reasonable extent, but for more details what makes PSC patients worry, why they may feel depressed or which activities they are unable to perform in particular, more disease and domain-specific questionnaires should be used for a full account of the relevant QoL issues in PSC. 36 QALY analysis is in general complicated by a heterogeneous and episodic disease course, which are both the case in PSC. 37 Taking into account these limitations, follow-up was split in different disease stages based on the available data. Recall bias can never be ruled out completely, although we deliberately chose in conjunction with the patient panel for 3-monthly questionnaires in order to limit this potential confounder as much as possible. Several studies have shown that a recall period of 3 months gives reliable data regarding medical consumption. 38,39 In addition, medical charts and annual letters to the GP were scanned for medical consumption items yearly.
In conclusion, these long-term FU data confirm that on population-based level LT-free median survival is more favourable than reported earlier. This study for the first time quantifies the severe burden PSC patients suffer from during their disease course, underscoring the urgent need for an effective therapy, and call for replication in other regions. Our data may serve as benchmark in (cost-)effectiveness analyses of new therapies, as well as may aid healthcare planners in future budget allocation.

FU N D I N G I N FO R M ATI O N
The study was funded by the Netherlands Organisation for Health Research and Development (ZonMw, grant number: 836041010).