Mortality attributable to cholestatic liver disease in the United States


  • Potential conflict of interest: Nothing to report.


In the past 2 decades, important advances have been made in the treatment of cholestatic liver diseases, including primary biliary cirrhosis (PBC) and primary sclerosing cholangitis (PSC). Whether these new therapies have had demonstrable impact on mortality on a population-wide scale has not been evaluated. This study describes the age-specific and sex-specific mortality rates from PBC and PSC in the United States between 1980 and 1998, based on the Multiple Cause of Death files. Age-specific and sex- specific mortality rates from PBC and PSC were calculated. The multivariable Poisson model was used to evaluate temporal changes in mortality rates. In 1998, the total age-adjusted and sex-adjusted PBC-related mortality rate was 0.24 per 100,000, and the age-adjusted and sex-adjusted PSC-related mortality rate was 0.23 per 100,000. During the observation period, PBC-related mortality significantly decreased over time in women younger than 65 years, and in men of all age groups, whereas in older women this number increased over time. PSC-related mortality remained essentially stable, except in men 65 years of age or older. Conclusion: Since the early 1980s, significant changes in mortality from PBC have occurred. The most noticeable change was an increase in the age of death, which indicates prolongation of survival. These changes may be attributable to liver transplantation or ursodeoxycholic acid. In contrast, mortality from PSC remained largely unchanged, highlighting the need for more effective therapeutic strategies. (HEPATOLOGY 2008.)

Primary biliary cirrhosis (PBC) and primary sclerosing cholangitis (PSC) are chronic cholestatic disorders of the liver. PBC is characterized histologically by portal inflammation and immune-mediated destruction of the intrahepatic bile ducts and serologically by the presence of antimitochondrial antibodies.1 PSC is characterized by progressive destruction of intrahepatic and extrahepatic bile ducts.2

PBC is a disease that affects primarily women, in a 9:1 ratio. The median age of diagnosis is 50 years (range, 20–90 years).1, 3 In the first population-based study in the United States, Kim et al.4 reported an incidence of PBC of 2.7 per 100,000 person-years and a prevalence of 40.2 per 100,000 persons. Based on these data, it was estimated that there would be 47,000 prevalent cases of PBC in the US white population and that approximately 3,500 new cases would be diagnosed each year.4 The overall median survival ranges between 10 and 15 years, with disease progression being more likely in patients with symptomatic disease, even though asymptomatic patients still have a lower overall median survival compared with an age-matched and sex-matched healthy population.1, 3 In a study by Prince et al.,5 only one-third of the deaths in asymptomatic patients with PBC were liver related. Several reports demonstrate an increased risk of hepatocellular carcinoma in patients with PBC; however, the crude frequency of hepatocellular carcinoma is still low at 0.7% to 3.6% (5.9%–11.1% in patients with advanced PBC).6–8

Although the incidence of PBC in the United States has not changed for the past 20 years,4 at least 2 therapeutic modalities have been shown to be effective in treating patients with PBC. These include orthotopic liver transplantation (OLT) and ursodeoxycholic acid (UDCA). Liver transplantation, shown to improve survival of patients with end-stage liver disease from PBC,9 has been used widely in PBC patients since the mid-1980s. UDCA therapy slows disease progression and reduces the need for liver transplantation in patients with PBC.10 It was approved in the United States for PBC in 1997, although its effectiveness had been reported in the literature as early as 1991.11 With these significant advances in the treatment of PBC, improvement in survival of patients with PBC is expected. However, such impact on PBC-related mortality has not been evaluated on a population-wide scale.

PSC affects men more commonly than women, and the median age of diagnosis is approximately 40 years. The median survival is 12 years from diagnosis. Its association with inflammatory bowel disease is well known.2 In a US population-based study, the estimated overall age-adjusted and sex- adjusted incidence of PSC was 0.90 per 100,000 person-years (1.25 in men and 0.54 in women), and the prevalence was 13.6 per 100,000 (20.9 men and 6.3 women). In this study, the estimated 10-year survival of patients with PSC was 65%, compared with an expected survival of 94%.12 Cholangiocarcinoma is the most feared complication and occurs in 7%–15% of patients with PSC, with an annual incidence of 0.5% to 1%.

Liver transplantation has achieved overall excellent outcomes, including long-term survival and improvement in quality of life in patients with advanced PSC.13 However, PSC tends to recur in the allograft, which sometimes leads to graft failure, retransplantation, or death.14 Furthermore, no medical treatment has been shown effective in slowing disease progression or time to liver transplantation in patients with PSC.15 In practice, however, UDCA may be being used off label, although its efficacy in patients with PSC has not been demonstrated.16

In this study, we used detailed death certificate data to describe the age-specific mortality rates from PBC and PSC in the United States between the years of 1980 and 1998 and to evaluate their time trends. The comparison between PBC and PSC provides helpful insight into the impact of advances in therapy for these diseases over time.


MRR, mortality rate ratio; OLT, orthotopic liver transplantation; PBC, primary biliary cirrhosis; PSC, primary sclerosing cholangitis; UDCA, ursodeoxycholic acid.

Patients and Methods

Data Source.

Data on mortality were obtained from the Multiple Cause of Death files, which comprise all death certificates issued in the United States (approximately 2 million annually). Data on all calendar years between 1980 and 1998 were obtained from the National Center for Health Statistics (Hyattsville, MD). Available data elements included age (in 5-year increments), sex, race (white, African, and other), and place of birth and death, and all listed (immediate, contributing, and underlying) causes of death on the death certificate.

The timeframe of analysis (1980–1998) was chosen because the diagnostic coding system used for the data was consistent during the period; the coding system was changed to the International Classification of Diseases 10th version (ICD-10) in 1999.

Case Ascertainment.

The Multiple Causes of Death files used the International Classification of Diseases 9th version (ICD-9) for diagnostic coding. Death records with a PBC diagnosis was identified with the code 5716 (biliary cirrhosis), whereas the code 5761 (sclerosing cholangitis) was used for PSC deaths. We considered a death PBC-related or PSC-related if either diagnosis was listed as the “underlying” or “immediate” cause of death or “underlying/immediate” cause of death was cirrhosis, portal hypertension and its complications (esophageal varices, ascites, hepatorenal syndrome, hepatic coma), hepatocellular carcinoma/cholangiocarcinoma, acute liver failure, or unspecified disorder of liver, and PBC or PSC was also listed as a contributing cause of death. Records in which PBC or PSC was listed but the immediate or underlying cause of death was not liver related were thought to represent deaths in which PBC or PSC was incidental and did not contribute to the death.

Data Analysis.

Data from the Multiple Cause of Death file were analyzed using SAS version 8.2 (SAS Institute Inc., Cary, NC). Mortality rates, namely incidence rates of deaths, attributable to PBC or PSC were calculated, considering the entire population of the United States to be at risk. Denominator age-specific, sex-specific, and race-specific person-years were based on the 2000 census. For the estimation of data variability, we assumed that, given a fixed number of person-years, the number of cases would follow a Poisson distribution, which allowed the estimation of standard errors and the calculation of 95% confidence intervals for the mortality rates. Because there were only a small number of decedents of nonwhite race, all races were combined in the estimation of mortality rates and subsequent analyses. Also, only those deaths in which the age was 20 years or older were included in the analysis.

The impact of age, sex, and calendar year was examined using the generalized linear model assuming a Poisson error structure. In the Poisson analysis modeling, both PBC-related death and PSC-related death, the natural log of the national population for a given age/sex combination was used as an offset. Thus, all models are age-adjusted and sex-adjusted to the national population. Interactions among these variables were examined to define whether there were differences in the temporal trend in incidence rates of PBC-related and PSC-related death across sex.


Deaths of PBC and PSC.

A total of 11,860 records of potential PBC-related deaths were found in the Multiple Cause of Death file, but only 7,913 (67%) met our definition of PBC-related death and were included in the analysis. Of those, 6,414 (81.1%) were women, and 1,499 (18.9%) were men. The mean age at death was 65.6 (±13.9) years. Most were of white race (93.7%). African-Americans accounted for 4.5%; others, which included Asian, Pacific Islander, Native American, and Alaskan, for 1.7% (Table 1).

Table 1. Mortality Attributable to PBC and PSC
  1. HCC, hepatocellular carcinoma; CCC, cholangiocarcinoma.

Number of all deaths11,86023,985
Number of deaths attributable to the disease7,913 (67%)8,081 (34%)
Mean age at death in years65.6 ± 13.968.8 ± 16.2
Age group  
 <2098 (1.2%)41 (0.5%)
 20–39146 (1.9%)361 (4.5%)
 40–642,459 (31.1%)1,992 (24.6%)
 ≥655,210 (65.8%)5,687 (70.4%)
 6,414 (81.1%)4,159 (51.5%)
 White7,417 (93.7%)7,145 (88.4%)
 African360 (4.5%)680 (8.4%)
Other (Asian, Pacific Islander, Native American, and Alaskan)136 (1.7%)256 (3.2%)
Concomitant cirrhosis/portal hypertension diagnosis6,020 (76.1%)3,901 (48.3%)
Concomitant HCC/CCC diagnosis230 (3%)1,002 (12.4%)

A total of 23,985 records of potential PSC-related deaths were found in the Multiple Cause of Death file, but only 8,081(34%) met our definition of PSC-related death and were included in the analysis. Of those, 4,159 (51.5%) were men, and 3,922 (48.5%) were women. The mean age at death was 68.8 (±16.2). Most were of white race (88.4%), African-Americans accounted for 8.4%, and others, which included Asian, Pacific Islander, Native American, and Alaskan, for 3.2% (Table 1).

The total age-adjusted PBC-related mortality in 1998 was 0.241 per 100,000. As expected, PBC mortality was much higher in women (0.365 per 100,000) than in men (0.082 per 100,000). In contrast, mortality attributable to PSC was higher in men (0.271 per 100,000) than in women (0.186 per 100,000), with the sex difference being much less dramatic with PSC than with PBC. The total age-adjusted PSC-related mortality in 1998 was 0.226 per 100,000.

Time Trends in PBC-Related and PSC-Related Mortality.

Figure 1 illustrates age-specific mortality for women and men for representative calendar years (1980, 1986, 1992, and 1998). In general, mortality from PBC increased with age in both sexes. In women, PBC mortality in the 1980s had a bimodal pattern, with a first peak in the late 50s and early 60s and the second rise in the 70s. Into the 1990s, the first peak disappeared, giving rise to a pattern of uniform increase with age. PBC mortality in men did not show a consistent pattern, in part because of the fewer number of deaths. Figure 2 displays the same information for PSC mortality. Mortality attributable to PSC also increased with age. There did not seem to be any longitudinal pattern to PSC mortality in men or women.

Figure 1.

PBC-related deaths for females and males. PBC, primary biliary cirrhosis.

Figure 2.

PSC-related deaths for females and malesw. PSC, primary sclerosing cholangitis.

Trend in mortality is further analyzed in Figs. 3 and 4, which show mortality rates by calendar year. Figure 3 demonstrates that in women between 40 and 64 years of age, death attributable to PBC decreased from 0.4 per 100,000 in 1980 to 0.2 per 100,000 in 1998. In contrast, in women in the older age group (>65 years), the rate increased over time from 0.93 per 100,000 in 1980 to 1.48 per 100,000 in 1998. In the youngest age group, the low mortality rate makes it difficult to appreciate a trend graphically; however, the Poisson model also showed a decrease over time (see later discussion). In men 40 to 64 years of age, the PBC death rate decreased from 0.1 per 100,000 in 1980 to 0.03 per 100,000 in 1998. Mortality rates also decreased in men of the other age groups.

Figure 3.

Time trend in PBC-related mortality. PBC, primary biliary cirrhosis.

Figure 4.

Time trend in PSC-related mortality. PSC, primary sclerosing cholangitis.

With regard to PSC, Fig. 4 shows that the death rate in men older than 65 years of age decreased over time from 1.5 per 100,000 in 1980 to 0.93 per 100,000 in 1998. In men of the other age groups, PSC mortality did not change over time. In women, there was no discernible trend in PSC mortality.

These findings were formally examined in Poisson regression models, shown in Tables 2 (PBC) and 3 (PSC). The mortality rate ratio (MRR) calculation showed that PBC mortality in women increased with age. For example, PBC mortality in women older than 65 years was 1.77 times higher than that in women aged 40 to 64 years (1.77 = 1/MRR, MRR = 0.56), and 41.7 times higher than that in women aged 20 to 39 years (41.7 = 1/MRR, MRR = 0.024). With regard to a time trend, in women older than 65 years of age, each decade in calendar year was associated with a 29% increase in PBC mortality, whereas in younger women, the mortality rate decreased by approximately half (MRR = 0.53). In men, as in women, PBC mortality increased with age. PBC mortality in men showed a significant reduction over time, regardless of their age, although the decrease was more pronounced in the 40 to 64 and 20 to 39 age groups (MRR = 0.42 and 0.48, respectively) than in the older counterpart (MRR = 0.78).

Table 2. PBC-Related Deaths
Mortality Rate Ratio [95% Confidence Interval]PMortality Rate Ratio [95% Confidence Interval]P
  1. NOTE. Calendar year is in decades. The MRR for the age groups, as well as the P-values, use the older age group as a reference group.

Age group 20–390.024 [0.019, 0.034]<.010.019 [0.011, 0.032]<.01
Age group 40–640.56 [0.51, 0.62]<.010.31 [0.25, 0.37]<.01
Age group 65 +1.001.00
20–39 Calendar year0.53 [0.37, 0.77]<.010.48 [0.26, 0.86]0.01
40–64 Calendar year0.53 [0.49, 0.58]<.010.42 [0.36, 0.50]<.01
65 + Calendar year1.29 [1.22, 1.36]<.010.78 [0.69, 0.87]<.01
Table 3. PSC-Related Deaths
Mortality Rate Ratio [95% Confidence Interval]PMortality Rate Ratio [95% Confidence Interval]P
  1. NOTE. Calendar year is in decades. The MRR for the age groups, as well as the P-values, use the older age group as a reference group.

Age group 20–390.026 [0.019, 0.036]<.010.022 [0.017, 0.029]<.01
Age group 40–640.17 [0.15, 0.20]<.010.16 [0.14, 0.18]<.01
Age group 65 +1.001.00
20–39 Calendar year0.73 [0.53, 1.01]0.061.12 [0.89, 1.42]0.34
40–64 Calendar year0.85 [0.75, 0.97]0.011.03 [0.93, 1.14]0.59
65 + Calendar year1.04 [0.98, 1.12]0.200.90 [0.84, 0.96]<.01

Mortality attributable to PSC also increased significantly with age in both men and women. Overall, there was much less noticeable calendar year effect in PSC mortality. The model did show that the PSC mortality decreased significantly in women between 40 and 64 years of age and men older than 65 years. However, the effect size was small (0.85 for the former and 0.90 for the latter).


Several epidemiological factors may contribute to changes in mortality rates of a given disease. First, if the disease incidence changes, morbidity and mortality secondary to the disease will be affected accordingly. In the case of PBC and PSC, there is no firm evidence that their incidence has changed over time.4, 12 Second, if there is a shift in disease severity affecting the case fatality rate, mortality rates in the population may change, even if the incidence remains the same. Third, similarly, if there is effective treatment that alters survival of patients, it may alter mortality rates. In the case of cholestatic liver diseases, therapeutic modalities that may have affected mortality rates on the population level include liver transplantation and UDCA.

Liver transplantation has been widely used since the mid-1980s, and it has been shown to be an effective treatment for end-stage liver disease caused by both PBC and PSC, with excellent outcomes reported. Survival after liver transplantation for PBC are on the order of 85%–90% and 70%–80% after 1 and 5 years,17 respectively, and for PSC, reported 1-year and 5-year survival rates are as high as 94% and 86%, respectively.18

UDCA has been shown in randomized controlled trials to delay disease progression and to improve survival free of liver transplantation in patients with PBC, especially for those started on treatment while early-stage disease is present.10, 19–24 UDCA also has been actively studied for the treatment of PSC; however, no effect on delaying disease progression or the need for liver transplantation has been shown,25 and further studies are ongoing, looking at the effects of high-dose UDCA in these patients.

Because the incidence of PBC in the United States has remained stable for the past 20 years,4 the trend observed in PBC-related mortality suggests that there was a beneficial effect of both OLT and UDCA. The mortality change observed in the mid-1980s is probably attributable to the more widespread use of liver transplantation. The stronger effect noted in younger women may be explained by the fact that younger patients are more likely to undergo liver transplantation. One may speculate that the increased mortality observed in older women could represent an effect of UDCA on delaying disease progression; however, because disease progresses at an older age, these patients may be less likely to benefit from liver transplantation. However, UDCA's effect might be more readily determined in a future study that looks at similar rates in the decade after its approval and more widespread use (1999–2009).

Interpretation of data on PSC is more difficult than that for PBC. In spite of the significant P-values in the Poisson models, inspection of Figs. 3 and 4 shows that the curves for PSC are largely flat over time. The less pronounced time trend on PSC-related mortality may represent the lack of medical treatment that is able to impact on disease progression. Another possible explanation is the fact that the death certificate data for sclerosing cholangitis include an unknown proportion of non-PSC cases. Potential scenarios of secondary sclerosing cholangitis that may be included in the data are cases of bile duct complications after surgery of the gallbladder or bile ducts and those of cholangiocarcinoma occurring in non-PSC patients. Because of the rapid adoption of laparoscopic surgery during the 1980s and 1990s, there may have been an increased incidence of secondary sclerosing cholangitis, which may in part explain the high proportion of PSC-related deaths in women, not in keeping with the male preponderance of the disease. Mortality from those patients may have obscured a reduction in mortality in patients with PSC, for example, secondary to OLT. Conversely, cholangiocarcinoma occurring in older patients without PSC may have been erroneously classified as PSC in the past, which may occur less frequently as diagnostic accuracy improves over time. This may have produced a trend in the reduction of mortality in the older age group.

One may note a noticeable discrepancy between the estimated number of incident cases (for example, 3,500 new PBC cases per year among white Americans4) versus the number of deaths reported here (for example, 470 PBC deaths per year on average). There may be several explanations for this gap between the incidence and mortality for PBC and PSC. First, there are patients with asymptomatic disease that remains undiagnosed at the time of death. The frequency with which this occurs may not be estimated based on currently available data. Second, it is well known that even patients with symptomatic PBC do not die of liver-related causes. According to Prince et al.,5 31% and 57% of deaths in PBC patients are liver related in asymptomatic and symptomatic patients, respectively. In the current data, 7,913 of 11,860 deaths (67%) in subjects with a PBC diagnosis were determined to be liver related. Clearly, not all patients with PBC die of it, and death certificates of decedents who had PBC but did not die of it may lack the PBC diagnosis. Third, the Olmsted County population from which the estimates for PBC and PSC incidence was derived was predominantly of Scandinavian descent, the geographic area with the highest reported incidence of PBC.26 Finally, the death certificate data entail inaccuracies, and relevant deaths may have been omitted or misclassified.27 As pointed out earlier, this bias is likely greater for PSC deaths. In summary, the impression that the death rate for PBC/PSC appears to be only a fraction of their estimated incidence is likely a result of several factors, including lack of diagnosis, deaths of causes unrelated to their liver disease, and geographic heterogeneity of the US population. It is likely that there is some underestimation in our data (that is, death certificate diagnosis) of the true mortality of liver-related causes with diagnosed PBC and PSC. The magnitude of this underestimation remains unknown, and it represents the main limitation of this study. However, it must be pointed out that the main focus of the study is the time trend in mortality, not necessarily the absolute number of deaths. Thus, as long as this underestimation remains consistent over time, we believe that our observation about the trend is valid.

The trend in the population-wide outcome of PBC and PSC observed in this study was recently corroborated by a study by Lee et al.,28 in which all patients who were registered or who underwent transplantation in the United States between 1995 and 2006 were analyzed. Their data are in support of this study in 2 ways. First, similar to our observation, there was a statistically significant decrease in the number of patients wait-listed for PBC over time, whereas no such trend was seen with PSC. Second, the average number of patients wait-listed for PBC and PSC in the United States was 368 and 465 per year, respectively. These numbers are similar to the number of deaths observed in this study, which indicate that our death certificate data are unlikely to be inaccurate by a large measure.

In conclusion, over a nearly 2-decade period, significant changes in mortality were observed, the most pronounced of which was the decreased mortality observed in younger women (<65 years old) with PBC. We postulate that these trends are attributable to advances in the treatment of cholestatic liver diseases with the advent of OLT and UDCA for the treatment of PBC. In contrast, the increased mortality in older women with PBC points to the fact that those treatment modalities are not curative. Mortality trends in PSC in general are less discernible, which may indicate lack of effective treatment options for patients with PSC.