Autoimmune, Cholestatic and Biliary Disease
Article first published online: 22 APR 2011
Copyright © 2011 American Association for the Study of Liver Diseases
Volume 53, Issue 5, pages 1590–1599, May 2011
How to Cite
Molodecky, N. A., Kareemi, H., Parab, R., Barkema, H. W., Quan, H., Myers, R. P. and Kaplan, G. G. (2011), Incidence of primary sclerosing cholangitis: A systematic review and meta-analysis. Hepatology, 53: 1590–1599. doi: 10.1002/hep.24247
Potential conflict of interest: Nothing to report.
Gilaad G. Kaplan is supported by a New Investigator Award from the Canadian Institute of Health Research and by a Population Health Investigator Award from the Alberta Heritage Foundation for Medical Research (which is now Alberta Innovates–Health Solutions). Natalie Molodecky is supported by a partnership studentship from the Alberta Heritage Foundation for Medical Research and the Crohn's and Colitis Foundation. Hude Quan is supported by a Health Scholar Award from Alberta Innovates–Health Solutions. Robert P. Myers is supported by a Clinical Investigator Award from Alberta Innovates–Health Solutions and a New Investigator Award from the Canadian Institute of Health Research. This research was funded by the Alberta Inflammatory Bowel Disease Consortium, which is supported by an interdisciplinary team grant from the Alberta Heritage Foundation for Medical Research.
- Issue published online: 22 APR 2011
- Article first published online: 22 APR 2011
- Accepted manuscript online: 23 FEB 2011 09:37AM EST
- Manuscript Accepted: 31 JAN 2011
- Manuscript Received: 29 NOV 2010
Incidence studies of primary sclerosing cholangitis (PSC) are important for describing the disease's burden and for shedding light on the disease's etiology. The purposes of this study were to conduct a systematic review of the incidence studies of PSC with a meta-analysis and to investigate possible geographic variations and temporal trends in the incidence of the disease. A systematic literature search of MEDLINE (1950-2010) and Embase (1980-2010) was conducted to identify studies investigating the incidence of PSC. The incidence of PSC was summarized with an incidence rate (IR) and 95% confidence intervals. The test of heterogeneity was performed with the Q statistic. Secondary variables extracted from the articles included the following: the method of case ascertainment, the country, the time period, the age, the male/female incidence rate ratio (IRR), and the incidence of PSC subtypes (small-duct or large-duct PSC and inflammatory bowel disease). Stratified and sensitivity analyses were performed to explore heterogeneity between studies and to assess effects of study quality. Time trends were used to explore differences in the incidence across time. The search retrieved 1669 potentially eligible citations; 8 studies met the inclusion criteria. According to a random-effects model, the pooled IR was 0.77 (0.45-1.09) per 100,000 person-years. However, significant heterogeneity was observed between studies (P < 0.001). Sensitivity analyses excluding non–population-based studies increased the overall IR to 1.00 (0.82-1.17) and eliminated the heterogeneity between studies (P = 0.08). The IRR for males versus females was 1.70 (1.34-2.07), and the median age was 41 years (35-47 years). All studies investigating time trends reported an overall increase in the incidence of PSC. Conclusion: The incidence of PSC is similar in North American and European countries and continues to increase over time. Incidence data from developing countries are lacking, and this limits our understanding of the global incidence of PSC. (HEPATOLOGY 2011;)
Primary sclerosing cholangitis (PSC) is a chronic cholestatic liver disease of an unknown etiology that is characterized by chronic inflammation, destruction, and fibrosis of the intrahepatic and/or extrahepatic biliary tree. This aberrant inflammatory response may ultimately lead to cirrhosis, end-stage liver disease, and the need for liver transplantation.1 Furthermore, PSC patients are at a considerably higher risk of developing colon cancer and cholangiocarcinoma.1-3 Studies have shown a strong association between PSC and inflammatory bowel diseases (IBDs); this is particularly true for ulcerative colitis.1 Approximately 2% to 7.5% of patients with ulcerative colitis have PSC, whereas 70% to 80% of PSC patients have ulcerative colitis or Crohn's disease.1, 4-6 The disease predominantly affects males,1 and the diagnosis is commonly made in the third or fourth decade of life.5 Epidemiological studies have reported that the incidence of PSC ranges from 0.04 to 1.30 per 100,000 person-years. Although the incidence of PSC is low, recent evidence suggests that it has increased in the last few decades.7-9
PSC poses a significant burden to the health care system because of its high mortality and morbidity rates and the need for liver transplantation.2, 10 Despite the severity of this condition, few population-based epidemiological studies have investigated the incidence of PSC.7-9, 11-15 Most studies that have described the epidemiology of PSC have used data from tertiary referral centers,16-20 and these data may be limited by a referral bias. Although several studies describing the incidence of PSC have been published, they have not been systematically summarized. The objectives of our study were to conduct a systematic review of the incidence of PSC with a meta-analysis and to provide recommendations for future studies describing the epidemiology of this disease. Insight into the incidence of PSC is important in describing the burden of the disease and may shed light on its etiology.
Patients and Methods
We conducted a systematic literature search with a predetermined protocol that was in accordance with the Meta-Analysis of Observational Studies in Epidemiology (MOOSE),21 which studied the quality of reporting.21 We searched MEDLINE (1950 to June 2010) and Embase (1980 to June 2010) for studies investigating the incidence of PSC. The search strategy is outlined in detail in Appendix I. The search was not limited by language or to human subjects. The reference lists of relevant articles were also reviewed.
Two reviewers (N.A.M. and H.K.) identified articles eligible for further review by performing an initial screening of identified abstracts and titles. Abstracts were eliminated if they were not observational and did not investigate the epidemiology of PSC. Studies that did not report original data (e.g., review articles) were also excluded. The full text of the remaining articles was retrieved and systematically reviewed according to the inclusion and exclusion criteria. Articles were included if they reported an incidence rate (IR) of PSC or enough information to calculate the IR. Disagreements between reviewers were resolved by consensus with third-party experts (R.P.M. and G.G.K.).
Two reviewers independently extracted data for each study. The variable of interest was the incidence of PSC. The IR per 100,000 person-years with 95% confidence intervals (CIs) was documented for the overall study period and for individual years when they were reported. Secondary variables extracted from the articles included the following: the method of case ascertainment (i.e., a patient registry or administrative database), the country of origin, the study time period, the median age and range, the male/female incidence rate ratio (IRR), the incidence of small-duct and large-duct PSC, the percentage of PSC cases with IBD, and information on key indicators of study quality from MOOSE.21
The incidence of PSC was summarized with an IR, which was defined as the number of cases in a population per 100,000 person-years at risk in the population. IRs adjusted for confounding factors were selected over unadjusted IRs. The standard errors (SEs) and 95% CIs for the IRs were estimated under the assumption of a Poisson distribution. The ratio of males to females was summarized with an IRR, which was defined as the IR of PSC in males over the IR of PSC in females. When the IRR was not reported but the number of male and female incident PSC cases and the total study population were included, the IRR was calculated under the assumption that the background population was 50% male. Heterogeneity was assessed with the Q statistic (5% level), and meta-analyses were performed with random-effects models because of the presence of heterogeneity between studies. Stratified analyses and meta-regression were performed according to the methods of case ascertainment (i.e., administrative data versus patient registry data) and the region of publication (i.e., North America versus Europe). The study time periods were not stratified because there was considerable overlap between them. Temporal trends were calculated with Joinpoint regression analysis,22 by which, through a series of permutations, tests were performed to assess whether the addition of joinpoints resulted in statistically significant linear changes in the direction or magnitude of the rates in comparison with a linear line. Two joinpoints at most were considered. The parameter estimate used to summarize the trend over the fixed interval was the average annual percentage change (AAPC) according to a generalized log-linear model that assumed a Poisson distribution. Sensitivity analyses were conducted by the exclusion of studies that were not population-based because this was considered the most important difference in the quality of the studies. The possibility of publication bias was assessed with the Begg test.
The search retrieved 718 and 951 citations from MEDLINE and Embase, respectively; 1607 of these citations were excluded after an initial screening based on titles and abstracts, and this left 62 articles for the full-text review (Fig. 1). The observed agreement between reviewers for the eligibility of articles during the initial screening was 92% (κ = 0.85). Upon the full-text review of the 62 articles, 54 were excluded for reasons listed in Fig. 1, and this left 8 studies for final inclusion in the systematic review.7-9, 11-15 The agreement between reviewers for the eligibility of articles was 100% (κ = 1). Characteristics of the eight included studies are shown in Table 1.
|Author||Publication Year||Country (Region)||Study Period||Source of Cases||Population- Based||Method of Case Ascertainment (Comments)||PSC Cases, n||Person- Years||Age at Diagnosis (Years), Median (Range)||Male/Female, n (%)/IRR||PSC IR (per 100,000 Person-Years)||IBD With PSC, n (%)||AAPC, % (95% CI)|
|Kaplan et al.11||2007||Canada (Calgary)||2000-2005||Patient registry||Yes||Retrospective cohort||49||5,562,605||41 (29-60)‡||27/22 (55)/ 1.24 (0.82-1.81)||0.92 (0.68-1.21)||36 (74)||6.0**,∥|
|Kingham et al.12||2004||United Kingdom (South Wales)||1984-2003||Patient registry||Yes||Prospective cohort||46||5,020,000||52 (11-80)††||33/20 (62)/ 1.65 (0.92-3.03)†,††||0.91 (0.67-1.22)||33 (62)††||N/A|
|Bambha et al.13||2003||United States (Olmsted County, MN)||1976-2000||Administrative database||Yes||Medical record linkage system||22||2,444,444||40 (34-50)‡||15/7 (68)/ 2.31 (0.89-6.71)||0.90 (0.56-1.36)||16 (73)||N/A|
|Berdal et al.15||1998||Norway (Akershus)||1985-1994||Administrative database||Yes||ICD codes (validated by a review of medical records)||12||1,800,000||43§ (32-64)||7/5 (58)/ 1.40 (0.38-5.59)†||0.7 (0.34-1.16)||N/A||N/A|
|Lindkvist et al.9¶||2010||Sweden (Vastra Gotaland)||1992-2005||Administrative database||Yes||ICD codes (validated by a review of medical records)||199||16,311,475||38.5 (18-77)||142/57 (71)/ 2.58 (1.88-3.57)||1.22 (1.06-1.40)||152 (76)||3.06∥ (0.01-6.20)|
|Boberg et al.14¶||1998||Norway (Aker University Hospital catchment area)||1986-1995||Patient registry||Yes||Prospective cohort||17||1,300,000||37 (14-67)||12/5 (71)/ 2.40 (0.79-8.70)†||1.3 (0.8-2.1)||12 (71)||N/A|
|Escorsell et al.7*||1994||Spain (nationwide)||1984-1988||Patient registry||No||Questionnaire (responses from 69.7% of centers in the region)||37||96,200,000||42.3§ (12-75)#||26/17 (60)#/ 1.53 (0.80-3.00)†||0.04 (0.03-0.05)||20 (47)#||27.2∥|
|Card et al.8*||2008||United Kingdom (nationwide)||1991-2001||Administrative database||No||GPRD (not fully population-based and not validated for PSC)||149||36,341,463||55 (5-94)||1.85 (1.33-2.59)||0.41 (0.34-0.48)||N/A||4.1|
Epidemiology and Demographics.
The eight studies identified from the literature search that met our inclusion criteria were pooled to give an overall IR estimate of 0.77 (0.45-1.09) per 100,000 person-years at risk (Fig. 2). Statistically significant heterogeneity was observed between studies (Q statistic = 403.53, P ≤ 0.001). Two studies reported the incidence of large-duct PSC versus small-duct PSC. Kaplan et al.11 found a 5-fold higher rate, whereas Lindkvist et al.9 found a 9-fold higher rate of large-duct PSC versus small-duct PSC. No evidence of publication bias was found (Begg test: z = −1.12, P = 0.262).
The proportion of male incident PSC cases versus female incident PSC cases was reported in all eight studies. The IRR for males versus females was pooled to give an overall IRR estimate of 1.70 (1.34-2.07; Fig. 3). When we analyzed only those studies that reported IRRs without the assumption of a 50% male background population, the pooled IRR was 1.84 (1.18-2.51). In eight studies that reported the age at diagnosis, the pooled median age was 41 years (range = 35-47 years; Fig. 4). Six studies reported the proportion of incident PSC cases with a diagnosis of IBD. When these were pooled, the proportion of IBD in PSC cases was 68% (58%-77%; Fig. 5).
The source of case ascertainment varied between studies, with four studies using administrative databases and four using patient registries (Table 1). The IR of PSC did not significantly differ between the two methods when a meta-regression was performed (P = 0.845; Table 2). The eight included studies were conducted in North America and Europe. The two studies from North America11, 13 had similar estimates and gave a pooled IR of 0.91 (0.69-1.14) per 100,000 person-years at risk. Studies from Europe9, 12, 14, 15 had a lower pooled estimate of 0.72 (0.36-1.08); however, meta-regression analysis revealed no statistically significant difference between regions (P = 0.636).
|Analysis||Characteristic||Studies, n||Measure||Meta-Regression P Value|
|Epidemiology||IR||8||IR (95% CI): 0.77 (0.45-1.09)||N/A|
|Demographics||Male/female ratio||8||IRR (95% CI): 1.70 (1.34-2.07)|
|Male/female ratio*||4||IRR (95% CI): 1.84 (1.18-2.51)|
|Age||8||Median (range): 41 years (35-47 years)|
|IBD||6||% (SE): 67 (58-77)|
|Method of case ascertainment||Administrative database||4||IR (95% CI): 0.76 (0.11-1.41)||0.845|
|Patient registry||4||IR (95% CI): 0.81 (0.30-1.31)|
|Region of study||North America||2||IR (95% CI): 0.91 (0.69-1.14)||0.636|
|Europe||6||IR (95% CI): 0.72 (0.36-1.08)|
|Epidemiology||IR||6||IR (95% CI): 1.00 (0.82-1.17)||N/A|
|Demographics||Male/female ratio||6||IRR (95% CI): 1.77 (1.15-2.38)|
|Male/female ratio*||3||IRR (95% CI): 1.91 (0.79-3.03)|
|Age||6||Median (range): 41 years (35-47 years)|
|IBD||5||% (SE): 74 (69-79)|
|Method of case ascertainment||Administrative database||3||IR (95% CI): 0.98 (0.65-1.32)||0.914|
|Patient registry||3||IR (95% CI): 0.95 (0.76-1.13)|
|Region of study||North America||2||IR (95% CI): 0.91 (0.69-1.14)||0.581|
|Europe||4||IR (95% CI): 1.03 (0.77-1.29)|
Temporal Trends in PSC Incidence.
Five studies investigated temporal trends in PSC incidence7-9, 11 (Table 2). Four studies reported estimates for the trends; three of these demonstrated statistically significant increases at the 5% level [AAPC = 6.0%11 (unpublished data), AAPC = 27.2%,7 and AAPC = 3.1%9]. In the last study, a significant increase of 35.1% over a 10-year period was reported (P = 0.05). Another study reported a tendency toward increasing incidence; however, a statistically significant result was not found (AAPC = 4.1%).8 The study that did not report an estimate for the trend found a significant trend toward increasing incidence in men but not women (P < 0.01 and P = 0.6, respectively) when the overall study time period was considered.13 One study reported time trends for different subtypes of PSC but failed to find a statistically significant increase when either small-duct or large-duct PSC or PSC with or without IBD was considered.9
The exclusion of the two studies that were not fully population-based increased the pooled IR of PSC to 1.00 (0.82-1.17) per 100,000 person-years at risk (Fig. 6). When only these six studies were considered, statistically significant heterogeneity was not observed (Q statistic = 9.72, P = 0.08). The pooled IRR for males versus females did not significantly change when these studies were excluded; the estimated value was 1.77 (1.15-2.38). The median age remained the same; however, higher pooled estimates were found for the different methods of case ascertainment and the study regions (Table 2).
PSC is a rare disease of unknown etiology. Despite its low prevalence, the burden of disease is substantial because of the lack of effective therapeutic options and the high rate of complications, which predominantly affect young patients. Few population-based epidemiological studies have investigated the incidence of PSC, and as a result, the epidemiology of this disease remains poorly defined. Here we present a comprehensive overview of the incidence of PSC. The overall incidence of PSC was 0.77 per 100,000 person-years at risk. The incidence was largely unchanged in multiple stratified analyses exploring study characteristics (e.g., case ascertainment). The median age at the diagnosis of PSC was 41 years, with males having a nearly 2-fold greater risk of developing PSC versus females. The pooled proportion of IBD in PSC cases was 67%, and this was consistent with previous reports. Sensitivity analysis considering only population-based studies increased the IR estimate to 1.00 per 100,000 person-years at risk. Because population-based studies provide a more accurate and reliable estimate of the rate of disease, the IR of 1.00 per 100,000 is likely more representative of the true incidence of PSC.
Heterogeneity was observed between studies exploring the incidence of PSC. This heterogeneity may be explained by differences in the geographic region, differences in the study design (e.g., the method of case ascertainment), and intrinsic biases associated with observational studies. A stratified analysis was conducted to assess IRs in different geographic regions. All the studies originated from either North America or Europe. When stratification was performed by continent, no differences in IRs were found. Regional similarities in IRs may be due to the fact that the studies originated from countries that were comparable in ethnicity, the prevalence of IBD, and the rates of IBD susceptibility genes.23, 24 Data were lacking for regions of low IBD prevalence (e.g., the developing world), so we could not explore the incidence of PSC in these areas. Because the incidence and prevalence of IBD are highest in North America25, 26 and Europe,27 PSC estimates from these studies may overestimate the global health burden. Additionally, we explored whether the method of case ascertainment contributed to the heterogeneity observed between studies. IRs did not differ between studies using administrative databases and studies using patient registries, and this indicated the robustness of these incidence estimates. Ideally, we would have explored whether the year of study contributed to the heterogeneity observed between studies; however, the considerable overlap of the time periods prevented this analysis. Additionally, we would have investigated the differences in the incidence of small-duct PSC versus large-duct PSC; however, only two studies stratified their results by the different types of PSC. Differentiating PSC subtypes is important because small-duct PSC has a more benign prognosis than large-duct PSC.11, 28 Thus, future population-based studies are required to characterize the incidence of the different subtypes of PSC.
Sensitivity analysis excluding studies that were not population-based revealed no statistically significant heterogeneity. The inconsistency of heterogeneity was likely due to the much smaller IR estimates and the larger overall study populations in two of the studies that were not population-based. The study by Escorsell et al.7 ascertained cases through a questionnaire circulated to gastroenterologists and hepatologists in 33 hospitals throughout Spain; however, only 69.7% of the centers responded. Because of the non–population-based nature of this study and the common underreporting by physicians, the IR was likely largely underestimated. The study by Card et al.8 ascertained cases through the General Practice Research Database (GPRD). The GPRD is not population-based; therefore, its incidence estimate represents that of the GPRD and not the general population. Moreover, the code for PSC was not validated, and this poses additional challenges to its validity. These two studies were not fully comparable with the other studies included in the analysis.
The incidence of PSC appears to be increasing; however, additional studies are necessary to confirm this observation. This increase in PSC incidence may be a direct result of its link to IBD because recent evidence suggests that the incidence of IBD is still increasing in many regions of the world.26, 29-32 Studies investigating the incidence of PSC with and without IBD may help to resolve this issue; however, the decrease in power when these analyses are conducted may hinder the finding of statistical evidence. One study reported time trends in PSC with and without IBD but failed to find a statistically significant increase.9 Additionally, the observed increase may be due to improvements in the diagnostic abilities of physicians and diagnostic tools such as noninvasive imaging (i.e., magnetic resonance cholangiopancreatography).33, 34 Magnetic resonance cholangiopancreatography permits fast and highly accurate imaging of the biliary tree and has been used with increasing frequency as a noninvasive alternative to endoscopic retrograde cholangiopancreatography.33, 35-37 Furthermore, the increasing use of biological therapies (e.g., infliximab) and immunosuppressants (e.g., azathioprine and methotrexate), particularly in those with IBD,38-40 may have contributed to the greater detection of PSC over time. Biologics and immunosuppressants have been associated with drug-induced liver complications, hepatobiliary disease, and liver toxicity41, 42; therefore, the routine screening of liver function profiles for individuals taking these therapies has increased. Studies investigating the incidence of PSC should consider assessments of diagnostic tool utilization over time. Moreover, this increase in incidence may be a result of biases in observational studies. In particular, the study by Escorsell et al.7 had an unrealistically high AAPC of 27%. Many factors likely contributed to the obvious bias of this estimate. The population estimate used to calculate the incidence was taken from the end of the study period. An increase in the study population (i.e., 12 regions in Spain)43 over the time interval would have overestimated the increase in IR. Additionally, because of the retrospective nature of the study and the need for physicians to respond to a questionnaire, the response rate could have been lower for earlier time periods. Finally, detection bias could have been more pronounced at the end of the study period because of the increased availability and use of diagnostic technologies.
The relative paucity of incidence studies of PSC was likely secondary to the difficulty of ascertaining cases of PSC, particularly from administrative databases. Administrative databases have the potential to introduce misclassification errors for rare diseases such as PSC. For example, PSC does not have a distinct International Classification of Diseases (ICD) code (ninth or tenth revision) and instead is listed under cholangitis, which includes much more common acute conditions such as ascending cholangitis. This leads to the incorrect classification of PSC incidence with administrative databases in which ICD coding is used without validation.
Limitations of our systematic review should be considered. First, the number of studies included in the stratified analyses was relatively small, so the incidence of PSC in these strata may not be accurately represented. Second, the quality of the studies was not always optimal; this was shown by the inconsistent methods of case ascertainment. Third, because of the lack of data provided by each study for comprehensively studying the demographics and time trends of the incidence of PSC, secondary calculations were required. Fourth, our systematic review was limited to incidence data and not to prevalence data. Although prevalence may be helpful in describing the disease burden, it is a static measure of the proportion of PSC cases in a population and is, therefore, influenced by mortality. Because patients with PSC have a high mortality rate, with survival rates likely differing by the population, our interest was in summarizing the rate at which new PSC cases occurred. Finally, the results of the meta-analysis should be interpreted with caution because data pooling does not address the intrinsic biases of observational studies. Despite these limitations, this review provides a comprehensive summary of the current literature. The meta-analysis identified important deficiencies in the literature, so future studies should be conducted to address the paucity of data as well as study design and quality issues.
The objective of this review was to help us to estimate the public health burden of PSC; the meta-analysis demonstrated that the IR of PSC was 0.77 per 100,000 person-years at risk with a slightly higher estimate of 1.00 per 100,000 person-years when only population-based studies were considered. We feel that because of the increased quality of population-based studies, the latter estimate better reflects the true incidence of PSC. Additionally, the meta-analysis identified important study limitations; thus, future studies should be properly designed with high-quality and systematic methods of case ascertainment and should explore the incidence of both small-duct PSC and large-duct PSC. Furthermore, additional studies need to evaluate whether the incidence of PSC is truly increasing by analyzing the incidence of PSC with and without IBD as well as the utilization of diagnostic tools concurrently with the incidence of PSC. Finally, the meta-analysis highlights that researchers should continue to explore the incidence of PSC, particularly in the developing world.
All the researchers thank the Alberta IBD Consortium for its support.
- 11The burden of large and small duct primary sclerosing cholangitis in adults and children: a population-based analysis. Am J Gastroenterol 2007; 102: 1042-1049., , , , .Direct Link:
- 19Prevalence of autoimmune liver disease in Alaska natives. Am J Gastroenterol 2002; 97: 2402-2407., , , , , .Direct Link:
- 20Variant forms of autoimmune liver diseases. Gastroenterol Pol 2003; 10: 359-364., , .
- 22National Cancer Institute. Surveillance research. http://surveillance.can cer.gov/joinpoint. Accessed January 2011.
- 26The epidemiology of inflammatory bowel disease in Canada: a population-based study. Am J Gastroenterol 2006; 101: 1559-1568., , , , , , et al.Direct Link:
- 43World Bank. World development indicators database. http://data.world bank.org/data-catalog. Accessed January 2011.