Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
Potential conflict of interest: Nothing to report.
The recent epidemiology and outcomes of primary biliary cirrhosis (PBC) in North America are incompletely described, partly due to difficulties in case ascertainment. In light of their availability, broad coverage, and limited expense, administrative databases may facilitate such investigations. We used population-based administrative data (inpatient, ambulatory care, and physician billing databases) and a validated International Classification of Diseases coding algorithm to describe the epidemiology and natural history of PBC in the Calgary Health Region (population ≈1.1 million). Between 1996 and 2002, the overall age/sex-adjusted annual incidence of PBC was 30.3 cases per million (48.4 per million in women, 10.4 per million in men). Although the incidence remained stable, the prevalence increased from 100 per million in 1996 to 227 per million in 2002 (P < 0.0005). Among 137 incident cases with a total follow-up of 801 person-years from diagnosis (median 5.8 years), 27 patients (20%) died and six (4.4%) underwent liver transplantation. The estimated 10-year probabilities of survival, liver transplantation, and transplant-free survival were 73% (95% confidence interval [CI] 60%–83%), 6% (95% CI 2.5%–12.6%), and 68% (95% CI 55%–78%), respectively. Survival in PBC patients was significantly lower than that of the age/sex-matched Canadian population (standardized mortality ratio 2.87; 95% CI 1.89–4.17); male sex (hazard ratio [HR] 3.80; 95% CI 1.85–7.82) and an older age at diagnosis (HR per additional year, 1.06; 95% CI 1.03–1.10) were independent predictors of mortality. Conclusion: This population-based study demonstrates that the burden of PBC in Canada is high and growing. Survival of PBC patients is significantly lower than that of the general population, emphasizing the importance of developing new therapies for this condition. (HEPATOLOGY 2009.)
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Primary biliary cirrhosis (PBC) is a chronic cholestatic disorder characterized by nonsuppurative destruction of the interlobular and septal bile ducts that may lead to cirrhosis.1 The hallmark serologic feature is the presence of antimitochondrial antibodies (AMA).2 Since the first descriptive epidemiologic study published nearly 35 years ago,3 a wealth of literature has examined the epidemiology of PBC.4–12 PBC is generally considered a rare disease affecting predominantly women. However, incidence and prevalence figures have varied widely from two to 49 cases per million and 19 to 402 cases per million, respectively.4, 5 This variation may be attributable to differences in study populations and periods and the methods of case ascertainment. Typically, multiple case-finding approaches have been used, including surveys, laboratory reports, liver histology databases, transplant registries, and death certificates.4, 5 Although these multifaceted approaches are likely the most sensitive, the cost, complexity, and limited availability of these data sources are important obstacles to their widespread application. These difficulties may partly explain the relative paucity of North American epidemiologic studies of PBC; only four population-based studies, all of which are now small and/or outdated, have been reported.6–9 Because health care delivery in North America tends to be fragmented, particularly in the United States, collaboration between providers from diverse settings has been hindered.8
An alternative data source that may overcome some of these barriers is administrative data.13 Administrative databases are primarily maintained for reimbursement and management purposes but have been used increasingly for research applications in various fields, including hepatobiliary disease.14–17 Their widespread availability, broad geographic coverage, relatively complete capture of health care encounters, and low cost are important advantages over other data sources.13 In addition, because they are ubiquitous, administrative databases may facilitate comparisons of diseases across regions with variable access to other data sources.
Based on gaps in the existing PBC literature, we undertook this study to provide updated information regarding its epidemiology in a well-defined, North American population. To do so, we identified cases using population-based administrative databases and a diagnostic coding algorithm developed and validated in our health region.18 We aimed to determine whether the burden of PBC has risen over recent years, as suggested by some reports.10–12 In addition, our administrative database linkage system permitted investigation of the natural history of PBC from a population-based perspective. Such descriptive epidemiologic studies are essential for resource planning, patient counseling, and treatment decisions, and may further our understanding of the environmental and genetic influences on the pathogenesis of PBC.
AHCIP, Alberta Health Care Insurance Plan; CHR, Calgary Health Region; CI, confidence interval; HR, hazard ratio; IQR, interquartile range; PBC, primary biliary cirrhosis; PPV, positive predictive value; SMR, standardized mortality ratio; UDCA, ursodeoxycholic acid.
Patients and Methods
Data Sources and Study Population.
The study population included adults (≥20 years) with PBC in the Calgary Health Region (CHR) between fiscal years 1994 and 2003 (April 1, 1994, to March 31, 2004). The CHR is one of the largest fully integrated, publicly funded health care systems in Canada, and provides all medical and surgical care to residents of Calgary and surrounding communities in southern Alberta (population ≈1.1 million). The region includes 12 academic and community hospitals.
Three databases were used to identify potential PBC cases. The Physician Claims Database records claims submitted by Alberta physicians for services provided to registrants of the Alberta Health Care Insurance Plan (AHCIP), which includes 99% of Albertans. Approximately 4,500 providers submit over 36 million claims annually.19 The Inpatient Discharge Abstract Database contains demographic, diagnosis, procedure, and mortality information on all discharges from hospitals with the CHR. The Ambulatory Care Classification System Database contains information on ambulatory care data, including clinic and emergency department visits, same-day surgery, and outpatient procedures.19 These databases have been used to examine the epidemiology, outcomes, and coding accuracy of a variety of conditions.16, 18, 20
Administrative Data Case Definition.
PBC cases were identified using an administrative data algorithm requiring at least two health care contacts with a diagnosis code for PBC (ICD-9-CM 571.6 and ICD-10 K74.3) during the study interval. This algorithm was validated through a review of the medical records of 189 CHR patients with at least one encounter in the aforementioned databases, including a diagnosis code for PBC.18 Clinical (i.e., symptoms and ursodeoxycholic acid treatment), laboratory, and histologic stage21 at the time of PBC diagnosis were extracted from medical records by a trained physician using a structured data collection instrument. This definition had a positive predictive value (PPV) of 73% (95% confidence interval [CI] 61%–75%) for definite or probable PBC, defined as the presence of at least two of the following: (1) cholestatic liver biochemistry (i.e., elevated serum alkaline phosphatase and/or gamma glutamyl transpeptidase concentration); (2) positivity for AMA (titer ≥1:40) and/or antibodies against the pyruvate dehydrogenase complex;2 and (3) compatible liver histology.18 When applied to a cohort of 17 well-characterized patients from two clinical trials for PBC at the University of Calgary,22, 23 the algorithm's sensitivity was 94% (95% CI 71%–100%). Its PPV was 89% (95% CI 82–94%) for suspected cases of PBC, in whom the diagnosis was recorded in the medical record but insufficient confirmatory data was available (i.e. liver biochemistry, AMA serology, and liver histology).18 A recent epidemiologic study of PBC using the United Kingdom General Research Practice Database employed a similar definition.24
The incidence of PBC in the CHR was evaluated between fiscal years 1996 and 2002. We used a washout period of 2 years at the beginning of the study (1994 and 1995) to avoid including prevalent cases. The point prevalence on March 31 of each year was also determined. Nonresidents of the CHR and nonregistrants in the AHCIP were excluded. In calculating incidence and prevalence rates, the end-of-fiscal year population of the CHR was considered at risk.25 Overall rates were directly age/sex-adjusted to the 2001 Canadian population.26 Temporal trends were evaluated using generalized linear models assuming a Poisson error structure.27 Descriptive statistical methods were used to describe the characteristics of the cohort, and comparisons between groups were made using Fisher's exact and Kruskal-Wallis rank tests.
The primary outcome measure was all-cause mortality obtained from the Vital Statistics database, which contains death certificate–verified data regarding the underlying cause of death for CHR residents who died in Alberta.19 Secondary outcomes included transplantation (obtained from the University of Alberta Liver Transplant Database)28 and transplant-free survival. These outcomes were examined using Kaplan-Meier survival analyses with comparisons made using the log-rank test.29 The date of the first PBC contact in the administrative data was considered the date of diagnosis. The median interval between a clinical diagnosis of PBC and the first administrative data contact in our validation study was less than 2 months.18 Censoring occurred upon deregistration from the AHCIP or March 31, 2007, whichever came first. The cumulative survival of PBC patients was compared with the expected survival of the age/sex-matched Canadian population using the standardized mortality ratio (SMR). The SMR, which reflects the ratio of the observed to expected number of deaths, was calculated via the Ederer II method30 using conditional probabilities of death for the 2000 Canadian population.31 The impact of age, sex, and year of diagnosis (1996–1999 versus 2000–2002) on outcomes were examined using Cox proportional hazards regression.32 Exploratory analyses also examined the impact of clinical parameters (e.g., treatment, laboratory data, and stage) on survival in patients with complete clinical data. Log-log plots of survival confirmed satisfaction of the proportional hazards assumption (data not shown).
Statistical analyses were performed using Stata/IC version 10.0 software (StataCorp, College Station, TX) and SAS version 9.1.3 software (SAS Institute, Cary, NC) software. The protocol was approved by the Conjoint Health Research Ethics Board at the University of Calgary.
Between April 1, 1996, and March 31, 2003, 137 incident cases of PBC were identified in the CHR; 224 prevalent cases resided in the region at some point during this interval. The numbers of incident PBC cases by age and sex are illustrated in Fig. 1. The majority (83%) were women, and the median age at diagnosis was 53 years (interquartile range [IQR] 44–64 years). Fatigue and/or pruritus were present in 60% of patients (n = 78 with complete data). Ninety-one percent of patients with treatment data (67/74) received ursodeoxycholic acid (UDCA) at a median daily dosage of 1,000 mg (IQR 750–1000 mg [n = 65]) or 14.0 mg/kg (IQR 12.3–15.1 [n = 58]). Additional laboratory and histologic parameters are listed in Table 1.
Table 1. Characteristics of Patients Diagnosed with PBC in the Calgary Health Region Between 1996 and 2002 at the Time of Diagnosis
The overall age/sex-adjusted annual incidence of PBC between 1996 and 2002 was 30.3 per million. The incidence did not change during the 7-year study period (Poisson P = 0.89; Fig. 2), nor did the age (P = 0.25) or sex (P = 0.75) distributions of incident cases. Women were nearly five times as likely to be diagnosed with PBC as men. Age-adjusted incidence rates were 48.4 per million in women versus 10.4 per million in men (incidence rate ratio 4.83; 95% CI 4.76–4.90). The incidence of PBC was also highly dependent on age (Table 2). The highest adjusted incidence was observed among 60- to 79-year-old patients (63.0 per million; incidence rate ratio versus 20- to 39-year-old patients, 8.27 [95% CI 8.13–8.42]).
Table 2. Incidence and Point Prevalence Rates of PBC (per Million Population) According to Age and Sex in the Calgary Health Region
Age-specific incidence and prevalence rates were unadjusted. Overall incidence and prevalence rates in these columns were age-adjusted to the 2001 Canadian population.
Age-specific incidence and prevalence rates were sex-adjusted to the 2001 Canadian population. Overall incidence and prevalence rates in these columns were age/sex-adjusted.
Age/sex-adjusted point prevalence rates of PBC (as of March 31 for each year) increased from 100 to 227 per million between 1996 and 2002 (Poisson P < 0.0005; Fig. 3). As expected, adjusted point prevalence rates (as of March 31, 2002) were much higher among women than men (383 versus 55 per million; Table 2). The highest prevalence of PBC was among individuals aged 60 to 79 years (573 per million versus 38 per million in the 20- to 39-year-old category; rate ratio 15.5 [95% CI 15.2–15.8]).
Natural History of PBC.
The 137 incident cases of PBC were followed for a total of 801 person-years from diagnosis. After a mean/median follow-up of 5.8 years (IQR 4.2–8.0 years; range 10 days to 10.9 years), six patients underwent transplantation (4.4%) and 27 patients died (20%). No patient died after transplantation. The cause of death, which was available in 89% (24/27) of these patients, was liver-related in 12 (50%) patients and due to malignancy in two (8%) patients (one each of pancreatic and laryngeal cancer), ischemic heart disease in two (8%) patients, septicemia in two (8%) patients, and miscellaneous causes in the remainder (n = 6). The proportion of liver-related deaths was higher among patients under 60 years of age (100% versus 37%; P = 0.04), but did not differ according to sex (53% in women versus 44% in men; P = 1.00).
Overall survival of PBC patients after diagnosis is illustrated in Fig. 4. The annual mortality rate was 3.4% (95% CI 2.3%–4.9%). Estimated 1-, 5-, and 10-year survival rates were 93% (95% CI 87%–96%), 83% (76%–89%), and 73% (60%–83%), respectively. Figure 5 illustrates overall survival of PBC patients compared with the age-matched Canadian population. Survival was significantly lower in men than women (log-rank P < 0.00005) and in both groups compared with that expected among the general population (overall SMR 2.87; 95% CI 1.89–4.17). In women, estimated 1-, 5-, and 10-year survival rates were 96% (95% CI 90%–98%), 87% (79%–92%), and 80% (64%–90%), respectively, corresponding to an SMR of 2.10 (95% CI 1.20–3.40) compared with the general female population. In men with PBC, estimated 1- and 5-year survival rates were 78% (95% CI 55%–90%) and 64% (40%–80%), respectively. At 10 years following their diagnosis, none of the male patients was alive, corresponding to an SMR of 6.16 (95% CI 3.08–11.02) compared with the general population.
Predictors of Survival.
In addition to sex, survival was significantly lower with an older age at diagnosis (log-rank P < 0.00005; Fig. 6). The year of diagnosis (1996–1999 versus 2000–2002) was not significant (P = 0.81). In a multivariate Cox proportional hazards analysis including sex and age, men (hazard ratio [HR] 5.06; 95% CI 2.29–11.17) and older patients at diagnosis (HR per additional year, 1.10; 95% CI 1.06–1.14) had a higher risk of death. In an exploratory analysis including 39 patients with histologic data, bridging fibrosis or cirrhosis was not a significant predictor of mortality (HR 0.85; 95% CI 0.09–8.23). However, several laboratory parameters indicative of advanced disease were predictive. Specifically, lower serum albumin (HR per 1-unit decrease, 1.17; 95% CI 1.06–1.28 [n = 77 patients with complete data]), and higher international normalized ratio (HR 57.5; 95% CI 4.6–726 [n = 80]), bilirubin (HR 1.01; 1.00–1.02 [n = 79]), and creatinine (HR 1.02; 1.00–1.04 [n = 61]) concentrations were associated with mortality. Although 16% of UDCA-treated patients died, none of the untreated patients died during the follow-up period (log-rank P = 0.27; HR could not be calculated). Due to the small number of deaths and limitations in the available data, we did not conduct a multivariate analysis including these clinical parameters.
Liver Transplantation and Transplant-Free Survival.
Six patients (4.4% of the total; 5.6% of patients ≤65 years) underwent transplantation and 33 (24%) died or underwent transplantation during follow-up. The 10-year estimated probability of transplantation was 5.7% (95% CI 2.5%–12.6%). In an analysis excluding patients over 65 years at diagnosis (n = 29), in whom transplantation is generally contraindicated, the 10-year probability of transplantation was 6.9% (95% CI 3.1%–15.0%). Estimated 1-, 5-, and 10-year probabilities of transplant-free survival were 92% (95% CI 86%–95%), 80% (72%–86%), and 68% (55%–78%), respectively. Neither age (P = 0.28) nor sex (P = 0.85) was significantly associated with transplantation in multivariate analysis. However, men (HR 3.80; 95% CI 1.85–7.82) and older patients (HR per additional year, 1.06; 95% CI 1.03–1.10) had a greater risk of the combined endpoint of death or liver transplantation.
In this study, we describe recent trends in the epidemiology and natural history of PBC in Canada. The major strength of our study is the ability to accurately define the numerator and denominator populations by using a validated definition for case ascertainment,18 and an administrative database linkage system that ensures complete capture of health care encounters in a population from clearly defined geographic boundaries. Our natural history study is strengthened by its population-based nature, which limits the selection bias inherent in studies restricted to a few health care providers or tertiary referral centers.
Previous epidemiologic studies of PBC, mostly from Europe, reported incidence rates of two to 49 cases per million and prevalence rates of 19 to 402 per million.4, 5 The figures from our study (2002 incidence and prevalence of 36 and 227 per million, respectively) are among the highest reported, and if projected nationally, correspond to at least 5,500 prevalent cases and 900 new cases diagnosed annually in Canada. Our study updates the limited North American data regarding the epidemiology of PBC.6–9 In two Canadian studies from the 1980s, incidence and prevalence rates of approximately 3 and 25 per million, respectively, were reported.6, 7 These figures are an order of magnitude lower than we observed, likely due to the use of insensitive case ascertainment methods, the requirement for histological confirmation,6 and the inclusion of only hospitalized cases.7 In an American study, Kim et al.8 identified 46 incident cases of PBC in Olmsted County between 1975 and 1995. The annual incidence of 27 per million is similar to that of our study, while the prevalence of 402 cases per million is the highest reported. In a study restricted to Alaskan natives, Hurlburt et al.9 identified 18 PBC patients between 1984 and 2000, corresponding to a prevalence of 160 per million.
Between 1996 and 2002, the incidence of PBC in our region was stable, whereas the point prevalence doubled. These findings confirm data from some, but not all investigators. For example, data from Scotland revealed a stable incidence between 1986 and 1996 (48–55 cases per million) but an increase in prevalence from 186 to 379 per million.10 In northeast England, the prevalence rose from 149 to 251 cases per million between 1984 and 1994; the incidence was stable.11 The local rise in prevalence may be attributable to greater disease awareness and testing. Earlier diagnosis more recently with prolonged survival due to lead time bias could also be implicated, although the stable age distribution of incident cases would argue against this hypothesis. Alternatively, recently diagnosed patients may truly have a better prognosis. Indeed, the median age of prevalent cases was greater in 2002 than in 1996 (57 versus 53 years). Recent American data support this hypothesis. For example, Lee and colleagues described a significant reduction in the number of transplantations for PBC, but an increase for other conditions.33 In our province, the number of transplantations for PBC remained stable between 1996 and 2002 (G. Meeberg, personal communication, August 2008) despite the increased prevalence that we observed. In another study, Kim and colleagues34 reported a reduction in PBC-related mortality in the United States between 1980 and 1988. The authors hypothesized that this decline may reflect greater use of transplantation (at least in the 1980s) and a beneficial effect of UDCA. Prior to the increase in prevalence that we observed, UDCA was in widespread use since pivotal studies demonstrated its efficacy.35, 36 Alternatively, the observed rise in prevalence may reflect net migration of prevalent PBC cases into our region (e.g., for specialist care). Finally, differential accuracy of the coding algorithm over time may have played a role. For example, awareness or appropriate use of the ICD codes for PBC may have improved over time. Although we cannot exclude this hypothesis, the algorithm's PPV was stable (e.g., for definite or probable PBC: 74% in 1996–1998 versus 80% in 1999–2002; P = 0.62), suggesting that reduced specificity more recently cannot be implicated.18
The remainder of our epidemiologic findings is largely confirmatory of other data. The risk of PBC was higher among women, although the female/male ratio that we observed (≈5:1) is lower than typically reported. This may relate to greater diagnostic misclassification in men, for whom the PPV of our case definition was lower (60% versus 94% in women).18 The burden of PBC was also age-dependent. Although typically considered a disease of middle age, as supported by the high number of cases in patients aged 40 to 59 years, the highest disease rates were observed in 60- to 79-year-old individuals due to the relatively smaller size of this population. In women of this age group, the incidence approached 1 in 10,000, and the prevalence was nearly 1 in 1,000.
We also describe the natural history of PBC in an unselected Canadian cohort. This is the first contemporary, population-based study of a true incident cohort of PBC patients followed beyond the 1990s. The estimated 10-year probabilities of survival, transplantation, and transplant-free survival were 73%, 6%, and 68%, respectively. These figures compare favorably to other natural history studies of PBC.37–39 Importantly, these studies were restricted to patients on UDCA, which may explain the increase in survival observed in these studies and our own compared with two other population-based natural history studies. In the northeast England cohort, 10-year survival was only ≈45%.40 However, only 37% of these patients were receiving UDCA treatment, and the median dosage (450 mg/day) was much lower than currently recommended.1 In Olmsted County, 10-year survival was 59%; only one-third of the follow-up in these patients occurred during UDCA treatment.8 Unfortunately, pharmacologic data is not available in the administrative databases that we examined, thus the proportion of patients receiving UDCA is unclear. However, among incident cases whose records we reviewed, 91% received UDCA.18 Another potential explanation for these survival differences relates to the variable recruitment periods of these studies (1975–1995 for Kim et al.8; 1987–1994 for Prince et al.40; and 1996–2002 for our study), which is pertinent because the prognosis of PBC has likely improved.33, 34 We did not observe a significant difference in survival according to the year of diagnosis; however, the interval that we studied is likely too small to properly address this issue.
In keeping with other reports,8, 24, 37–42 the survival of PBC patients was reduced compared with the general Canadian population. Poupon et al.38 reported that 10-year survival among UDCA-treated patients was slightly lower than that of the French population, a difference largely observed among cirrhotic patients. More recently, the same investigators reported that overall survival in UDCA-treated patients was similar to that of controls, although patients with advanced disease (stages 3 or 4) had higher risk.43 In a study by Pares and colleagues,39 survival among UDCA responders was similar to the general Spanish population, but nonresponders had increased mortality. Unfortunately, we could not examine the impact of histological stage or response to UDCA on the prognosis of our patients. However, we identified two independent predictors of mortality: older age at diagnosis and male sex. Although no patients under 40 years of age died, 10-year survival was 82% in 40- to 59-year-old patients (SMR ≈4) and 54% in patients aged 60 to 79 years (SMR ≈3). We also observed five-fold mortality in men, a finding that has been reported37 but not confirmed in most studies.38–40, 43, 44 We suspect that this discrepancy may relate to greater diagnostic misclassification or delayed diagnosis in men (and older patients). For example, men with primary sclerosing cholangitis and alcoholic cirrhosis—two conditions with a worse prognosis than PBC—were more frequently incorrectly labeled as PBC than were men in our validation study.18 Similarly, among patients who underwent a liver biopsy, bridging fibrosis or cirrhosis tended to be more common in men (50% versus 24% ; P = 0.21). Likewise, median albumin levels tended to be lower (35 [IQR 29–39] versus 37 [IQR 34–40]; P = 0.19) and bilirubin levels higher (16 [IQR 12–27] versus 11 [IQR 7–20]; P = 0.05) in men. Not surprisingly, patients with biochemical indicators of advanced liver disease had a higher risk of death.
Our study has several limitations. Most importantly, misclassification of individuals without PBC due to our reliance on administrative data may have affected our results. However, we expect these numbers to be small considering the high sensitivity and PPV (94% and 89%, respectively) of our algorithm.18 Second, without clinical data, we could not examine potentially important risk factors for PBC, including environmental exposures that may play a role in disease pathogenesis.45–47 Future studies will assess local geographic variation in the burden of PBC that would strengthen the case for environmental influences. Third, the applicability of our results to other regions warrants confirmation. Because we have demonstrated the feasibility of using administrative databases to describe PBC epidemiology and outcomes, similar studies in other areas should be conducted. Investigations of this kind may further our understanding of the interplay between environmental and genetic influences on disease pathogenesis. Finally, we lacked detailed clinical data on all patients to fully evaluate various prognostic factors, including treatment, histology, autoantibodies, and laboratory data (such as Mayo risk score).48
In conclusion, using population-based administrative data, we have described the recent epidemiology and natural history of PBC in southern Alberta. The observed incidence and prevalence are among the highest reported. Whereas the incidence was stable, the prevalence doubled between 1996 and 2002, perhaps due to an improved prognosis or earlier diagnosis during recent years. Patient survival is lower than that of the general Canadian population, emphasizing the importance of developing new therapies for this condition.
We thank Dr. Alaa Rostom for helpful comments regarding the manuscript and Dr. Vince Bain and Glenda Meeberg for providing data from the University of Alberta Liver Transplant Database.