Comorbidity and survival of Danish cirrhosis patients: A nationwide population-based cohort study

Authors

  • Peter Jepsen,

    Corresponding author
    1. Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark
    • Department of Clinical Epidemiology, Olof Palmes Allé 43-45, DK-8200 Aarhus N, Denmark
    Search for more papers by this author
    • fax: (45)-89-42-48-01.

  • Hendrik Vilstrup,

    1. Department of Medicine V (Hepatology and Gastroenterology), Aarhus University Hospital, Aarhus, Denmark
    Search for more papers by this author
  • Per Kragh Andersen,

    1. Department of Biostatistics, Institute of Public Health, University of Copenhagen, Copenhagen, Denmark
    Search for more papers by this author
  • Timothy L. Lash,

    1. Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark
    2. Department of Epidemiology, Boston University School of Public Health, Boston, MA
    Search for more papers by this author
  • Henrik Toft Sørensen

    1. Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark
    2. Department of Medicine V (Hepatology and Gastroenterology), Aarhus University Hospital, Aarhus, Denmark
    Search for more papers by this author

  • Potential conflict of interest: Nothing to report.

Abstract

Patients with liver cirrhosis have a high mortality, not just from cirrhosis-related causes, but also from other causes. This observation indicates that many patients with cirrhosis have other chronic diseases, yet the prognostic impact of comorbidities has not been examined. Using data from a nationwide Danish population-based hospital registry, we identified patients who were diagnosed with cirrhosis between 1995 and 2006 and computed their burden of comorbidity using the Charlson comorbidity index. We compared survival between comorbidity groups, adjusting for alcoholism, sex, age, and calendar period. We also examined the risks of cirrhosis-related and non–cirrhosis-related death using data from death certificates and identified a matched comparison cohort without cirrhosis from the Danish population. We included 14,976 cirrhosis patients, 38% of whom had one or more comorbidities. The overall 1-year survival probability was 65.5%; the 10-year survival probability was 21.5%. Compared with patients with a Charlson comorbidity index of 0, the mortality rate was increased 1.17-fold in patients with an index of 1 [95% confidence interval (CI), 1.11–1.23], 1.51-fold in patients with an index of 2 (95% CI, 1.42–1.62), and two-fold in patients with an index of 3 or higher (95% CI, 1.85–2.15). In the first year of follow-up, but not later, comorbidity increased the risk of cirrhosis-related death, and this was consistent with an apparent synergy between the cirrhosis and comorbidity effects on mortality in the same period. Conclusion: Our findings demonstrate that comorbidity is an important prognostic factor for patients with cirrhosis. Successful treatment of comorbid diseases in the first year after diagnosis may substantially reduce the mortality rate. (HEPATOLOGY 2008.)

Liver cirrhosis is a life-threatening chronic disorder with an incidence rate of more than 190/1,000,000 per year in Denmark.1 Most patients with cirrhosis are in their fifties or older at the time of diagnosis,1 so many may have other diseases that are unrelated to cirrhosis, i.e., comorbidities2 and hence an increased mortality from several chronic diseases.3 Furthermore, cirrhosis shares risk factors with other chronic diseases, such as alcoholism, smoking, malnutrition, and obesity.4–6 For example, alcoholism, the most prevalent cirrhosis risk factor in Denmark,7 is also a risk factor for cancer and stroke.8–10 Improvements in prophylaxis and treatment of cirrhosis complications have decreased the risk of dying from cirrhosis,11 but this reduction inevitably increases the risk of dying from other causes. Nonetheless, the prognostic importance of comorbidity in cirrhosis patients is largely unknown.12, 13 Accurate data on prognosis and prognostic factors are important for clinical decision making and patient counseling, for understanding the clinical course of cirrhosis, for the design and analysis of epidemiologic studies of patients with cirrhosis, and for health care policy making.14, 15 We examined the prognostic impact of comorbidity in a large nationwide population-based study of cirrhosis patients followed for as many as 12 years.

Abbreviation

CCI, Charlson comorbidity index; CI, confidence interval; ICD, International Classification of Diseases.

Patients and Methods

Data were prospectively collected from a Danish nationwide population-based hospital registry.16 Denmark has 5.3 million inhabitants, and the National Health Service provides free tax-supported health care for all.

Data Sources

The National Patient Registry contains data from hospital contacts. It was established in 1977, and data from all inpatient admissions to a nonpsychiatric hospital in Denmark have been recorded since then.17 Data from outpatient visits have been recorded since 1995. Each record includes the dates of admission and discharge (outpatients: first and last visit) and up to 20 diagnoses, one of which is designated as the primary diagnosis. All diagnoses were coded according to International Classification of Diseases (ICD) 8 until 1993 and ICD10 from 1994 onward. Diagnosis codes are recorded at discharge by the physician who discharges the patient.

The Cause of Death Registry was established in 1943. Whenever a Danish citizen dies, the attending physician must report the cause of death, and a chain of events leading to death can be described by specifying up to four diagnoses. In the Cause of Death Registry, causes of death are translated into ICD10 diagnoses. Registration is currently complete through 2001.

Denmark's Civil Registration System updates the vital status of all Danish citizens on a daily basis, including date of death or emigration. Individual-level data from the National Patient Registry, the Cause of Death Registry, and the Civil Registration System can be linked through the unique personal identification number,16 which embeds information on birth date and sex.

Information on Patients with Cirrhosis

We included all patients who received their first hospital diagnosis, whether primary or secondary, of alcoholic or unspecified cirrhosis (ICD8: 571.09, 571.92, 571.99; ICD10: K70.3, K74.6) during an inpatient admission or an outpatient visit between January 1, 1995, and August 31, 2006.

Comorbidity.

We used the cirrhosis patients' primary and secondary diagnoses from all inpatient hospitalizations in the 10 years preceding their first diagnosis to identify comorbidities. The severity of comorbidity was based on the widely used Charlson comorbidity index (CCI), which was originally developed to predict 1-year mortality in hospitalized medical patients.18 This scoring system assigns between 1 and 6 points to a range of diseases (1 point for myocardial infarction, congestive heart failure, peripheral arterial disease, cerebrovascular disease, dementia, chronic pulmonary disease, connective tissue disease, ulcer disease, mild liver disease, and diabetes without organ damage; 2 points for diabetes with organ damage, hemiplegia, severe renal disease, and nonmetastatic cancer; 3 points for severe liver disease; 6 points for metastatic cancer and human immunodeficiency virus infection), and the sum of points serves as a measure of the burden of comorbidity. We defined the comorbid diseases according to the ICD10 codes provided by Quan et al.19 and to the ICD8 codes that matched the ICD10 codes as closely as possible. All patients had liver disease, so mild liver disease and severe liver disease were not counted as comorbidities. Hepatocellular carcinoma, on the other hand, was counted as a comorbid cancer.

Alcoholism is a risk factor for several of the diseases in the CCI, and it is also a prognostic factor among cirrhosis patients.20, 21 Therefore, we identified those patients who had been given a diagnosis suggesting chronic alcohol abuse (ICD8: 291.xx, 303.xx except 303.90, 571.09, 571.10; ICD10: F10.x except F10.0 and F10.1, G31.2, K70.x) after an inpatient hospitalization in the decade preceding their cirrhosis diagnosis.

Cause of Death.

Based on data from the Cause of Death Registry, deaths were classified as cirrhosis-related or non–cirrhosis-related. Cirrhosis-related deaths were those with at least one of the following listed as the cause of death or as part of the events leading to death: cirrhosis (K70.3, K74.6), liver failure (K70.4, K72.x), portal hypertension (K76.6), hepatorenal syndrome (K76.7), or esophageal or gastric varices (I85.x, I86.x). All other deaths were classified as non–cirrhosis-related.

Statistical Analysis

All-Cause Mortality.

Cirrhosis patients were followed from when they were first discharged from hospital with a diagnosis of cirrhosis until death or censorship at emigration or on December 31, 2006. The outcome was time to death.

We classified patients according to CCI (0, 1, 2, or ≥3), and computed the survival probability for each comorbidity group using the Kaplan-Meier estimator. This estimator does not control for confounding, however, so we also computed direct adjusted survival curves to control for sex, age at diagnosis, calendar period of diagnosis (1995–1998, 1999–2002, or 2003–2006), and alcoholism.22, 23

We used Cox proportional hazards regression to estimate the association between comorbidity group and mortality rate (specifically, the hazard rate), adjusting for sex, age at diagnosis, calendar period of diagnosis, and alcoholism. Our preliminary analyses indicated that the hazard rates for men and women were nonproportional, as were the rates for alcohol abusers and individuals who did not abuse alcohol. However, a regression model stratified by sex and alcoholism and with adjustment for age and calendar period yielded almost exactly the same association between CCI and mortality rate as did a regression model with adjustment for all four variables. Therefore, we used the model without stratification. Furthermore, we found that substituting current age, as a time-dependent variable, for age at diagnosis did not affect the results of the regression model. CCI and calendar period were included in the regression model using indicator variables. Age was included in the regression model as a continuous variable, and we used the procedure described by Royston et al. to examine whether a first- or second-degree fractional polynomial transformation of age would yield a better-fitting regression model and thus reduce residual confounding.24 In a supplementary analysis, we substituted the individual comorbidities for the comorbidity groups in the final regression model.

We examined whether the association between comorbidity group and mortality rate was constant over calendar time and patient characteristics. This examination was done by repeating the regression model for patients included in each of the three calendar periods of diagnosis separately, for men and women separately, and for alcohol abusers and those who did not abuse alcohol separately.

Cirrhosis-Related and Non–Cirrhosis-Related Mortality.

In this analysis cirrhosis patients were censored on December 31, 2001, because cause of death registration was incomplete after 2001, and deaths were classified as cirrhosis-related or from other causes. We computed the cumulative incidence of cirrhosis-related death and death from other causes, taking into account that these were competing risks.25

We used competing risks regression to estimate the associations between comorbidity group and cumulative incidence of cirrhosis-related death and of not cirrhosis-related death.26 Our preliminary analyses indicated that the associations were stronger shortly after cirrhosis diagnosis than later, so we computed separate estimates of association (specifically, subdistribution hazard ratios) for the first year after cirrhosis diagnosis and for the remainder of follow-up. The estimates were controlled for gender, age, calendar period, and alcoholism, coded and included as they were in the analysis of all-cause mortality.

Synergy Between Cirrhosis and Comorbidity Effects.

We identified a comparison cohort without cirrhosis through the Danish Civil Registration System. This cohort consisted of 10 Danish citizens per cirrhosis patient, matched for sex and birth year but otherwise random, who were alive and without cirrhosis when their matching cirrhosis patient was included. Like the cirrhosis patients, these persons were classified using the CCI with liver diseases ignored and followed until death or censorship at emigration or on December 31, 2006. However, they were also censored in the event of a cirrhosis diagnosis during follow-up.

Using both cohorts, we estimated the synergy between the prognostic effects of cirrhosis and comorbidity. The synergy was computed as the interaction risk,27 with a bootstrap 95% confidence interval (CI),28 on 1-year cumulative mortality and on 5-year cumulative mortality among 1-year survivors. Cumulative mortality was based on the Kaplan-Meier estimator.

Results

We included 14,976 cirrhosis patients in the study, 9391 (63%) of whom died during follow-up. The median age at inclusion was 56 years, and 66% were men. Sixty-two percent had a CCI of 0, 21% had an index of 1, 10% had an index of 2, and 7% had an index of 3 or higher (Table 1). Older patients had more comorbidities than younger patients, men had more comorbidities than women, and patients diagnosed late in the study period had more comorbidities than patients diagnosed early in the study period (Table 1). The most common comorbidities were ulcer disease [n = 1869 (12%)], diabetes [n = 1524 (10%)], and cancer [n = 1016 (7%)]. Approximately half of the patients with a CCI of 3 or higher had cancer (Table 1).

Table 1. Characteristics of 14,976 Cirrhosis Patients
CharacteristicsCCI = 0CCI = 1CCI = 2CCI = 3+Total
n%n%n%n%n%
  • Patients are categorized by the Charlson comorbidity index (CCI) at the time of cirrhosis diagnosis. HIV indicates human immunodeficiency virus.

  • *

    Only the number and percentage of patients with the comorbidity are shown.

Age at cirrhosis diagnosis          
 0–493,2093570722213149094,21928
 50–593,283351,0523346731232235,03434
 60–691,888208982847231351343,60924
 70–7970684271326518270261,66811
 80+168211348568084463
Sex          
 Women3,196351,0403349833291285,02534
 Men6,058652,157671,00467732729,95166
Year of cirrhosis diagnosis          
 1995–19983,131349743046431271264,84032
 1999–20023,121341,1013449933341335,06234
 2003–20063,002321,1223553936411405,07434
Comorbidity not in CCI*          
 Alcoholism6,353692,31072993666376210,29369
Comorbidities in CCI*          
 Ulcer disease001,1453636024364361,86912
 Diabetes006512042228451441,52410
  Diabetes without organ damage006512029820260251,2098
  Diabetes with organ damage00001248191193152
 Cancer000051834498491,0167
  Nonmetastatic cancer000045830327327855
  Metastatic cancer000000108111081
 Chronic pulmonary disease004401423716258259356
 Congestive heart failure00276922215309308075
 Cerebrovascular disease003451117211241247585
 Myocardial infarction0012041017168163893
 Peripheral arterial disease001063775158153412
 Severe renal disease0000755142142171
 Dementia005825545251651
 Connective tissue disease005623223131191
 Hemiplegia000081202280.2
 HIV infection000000202200.1
Total9,2541003,1971001,5021001,02310014,976100

All-Cause Mortality.

The overall survival probability was 65.5% after 1 year (95% CI, 64.7%–66.2%), 37.5% after 5 years (95% CI, 36.7%–38.4%), and 21.5% after 10 years (95% CI, 20.5%–22.5%). Survival depended markedly on CCI, although the differences in survival were attenuated by adjusting for sex, age at diagnosis cubed, calendar period, and alcoholism (Fig. 1). Compared with patients with a CCI of 0, the adjusted mortality rate was increased 1.17-fold for patients with an index of 1 (95% CI, 1.11–1.23), 1.51-fold for patients with an index of 2 (95% CI, 1.42–1.62), and two-fold for patients with an index of 3 or higher (95% CI, 1.85–2.15) (Table 2). Male sex increased the mortality rate 1.23-fold (95% CI, 1.18–1.29) and alcoholism 1.36-fold (95% CI, 1.29–1.42), whereas calendar period had no effect. The mortality rate increased with age, but the fractional polynomial analysis indicated that age cubed yielded a better-fitting regression model. This better fit implied that a 10-year age difference was more important for the risk of death among elderly patients than among younger patients. The association between comorbidity and mortality was constant over calendar time and patient characteristics (Table 2). The analysis of individual comorbidities showed that most comorbidities in the CCI were associated with mortality, and that the 2-point comorbidities were more strongly associated with mortality than the 1-point comorbidities, but not as strongly as the 6-point comorbidities (Supplementary Table 1).

Figure 1.

Survival by CCI for the 14,976 included cirrhosis patients. Survival probabilities are shown as Kaplan-Meier survival curves (gray), which are not adjusted for confounding, and direct adjusted survival curves (black), which are adjusted for confounding by sex, age cubed, calendar period, and alcohol abuse.

Table 2. Association Between CCI and Mortality Rate Within Patient Groups
 nCCI = 0CCI = 1CCI = 2CCI = 3+
MRR95% CIMRR95% CIMRR95% CIMRR95% CI
  1. The strength of association is expressed as the MRR with the associated 95% CI obtained from a Cox regression model adjusted for sex, age cubed, calendar period, and alcohol abuse. MRR indicates mortality rate ratio.

Year of diagnosis         
 1995–19984,8401.01.15(1.06–1.25)1.45(1.31–1.62)1.93(1.69–2.21)
 1999–20025,0621.01.14(1.05–1.24)1.57(1.41–1.75)2.07(1.83–2.34)
 2003–20065,0741.01.26(1.14–1.39)1.53(1.35–1.74)1.99(1.74–2.28)
Sex         
 Women5,0251.01.20(1.09–1.31)1.49(1.32–1.68)2.05(1.78–2.36)
 Men9,9511.01.16(1.09–1.23)1.52(1.41–1.65)1.98(1.81–2.17)
Alcoholism         
 No4,6831.01.20(1.09–1.33)1.58(1.41–1.77)2.31(2.04–2.62)
 Yes10,2931.01.16(1.09–1.23)1.49(1.38–1.62)1.86(1.69–2.05)
Total14,9761.01.17(1.11–1.23)1.51(1.42–1.62)2.00(1.85–2.15)

Cirrhosis-Related and Non–Cirrhosis-Related Mortality.

We included 8599 cirrhosis patients diagnosed before December 31, 2001, 4298 (50%) of whom died during follow-up. Seventy-three percent of the deaths were cirrhosis-related, and the cumulative incidence of cirrhosis-related death exceeded the cumulative incidence of death from other causes in all comorbidity groups at all times. As would be expected, the risk of death from causes other than cirrhosis increased with the comorbidity level, but so did the risk of death from cirrhosis, albeit less clearly and only shortly after diagnosis (Fig. 2). After adjustment for confounders, the cumulative incidence of cirrhosis-related death increased with the comorbidity level during the first year of follow-up [subdistribution hazard ratio for CCI of 1, 1.04 (95% CI, 0.93–1.16); for index of 2, 1.20 (95% CI, 1.05–1.38); for index of 3 or higher, 1.37 (95% CI, 1.17–1.62)], but not later. The cumulative incidence of death from other causes increased with the comorbidity level throughout the follow-up period, although more so during the first year after cirrhosis diagnosis.

Figure 2.

Cumulative incidence of death from cirrhosis and from other causes by CCI for the 8599 cirrhosis patients diagnosed before December 31, 2001.

Synergy Between Cirrhosis and Comorbidity Effects.

Of the 149,760 members of the matched cohort without cirrhosis, 87% had a CCI of 0, 7% had an index of 1, 4% had an index of 2, and 2% had an index of 3 or higher. In the first year after cirrhosis diagnosis, but not later, the prognostic effects of cirrhosis and comorbidity were synergistic (that is, the joint effects of cirrhosis and comorbidity exceeded the sum of their individual effects on mortality); 14% (95% CI, 10.4%–17.4%) of cirrhosis patients with a CCI of 3 or higher died from this synergy (Table 3).

Table 3. Synergy Between Cirrhosis and Comorbidity Effects
 No Cirrhosis, No ComorbidityCirrhosis, No ComorbidityComorbidity DefinitionNo Cirrhosis, ComorbidityCirrhosis, ComorbidityInteraction Risk* (95% CI)
  • Each row shows the cumulative mortality in the four combinations of ±cirrhosis and ±comorbidity, but the definition of comorbidity differs between rows. The far right column shows the interaction risk (that is, the percentage of patients whose death was attributable to interaction between cirrhosis and comorbidity).

  • *

    Among cirrhosis patients with a CCI of 2, in the first year of follow-up, an estimated 28.3% (29.0–0.7) died from cirrhosis, 7.1% (7.8–0.7) died from comorbidity, and 11.0% (47.1–28.3–7.1–0.7) died from the synergy between cirrhosis and comorbidity.

  • Conditional on being alive at 1-year follow-up.

1-Year follow-up0.7%29.0%CCI = 13.9%37.0%4.9% (2.9% to 6.7%)
 0.7%29.0%CCI = 27.8%47.1%11.0% (8.2% to 13.7%)
 0.7%29.0%CCI = 3+18.8%61.0%14.0% (10.4% to 17.4%)
5-Year follow-up4.7%39.1%CCI = 118.2%46.4%−6.2% (−9.3% to −3.2%)
 4.7%39.1%CCI = 225.4%56.4%−3.5% (−8.3% to 1.0%)
 4.7%39.1%CCI = 3+41.9%69.3%−7.0% (−13.5% to −0.2%)

Discussion

In this nationwide population-based study, we followed nearly 15,000 cirrhosis patients from diagnosis and found that the burden of comorbidity at the time of cirrhosis diagnosis was strongly associated with mortality. This association was due in part to a higher risk of cirrhosis-related death for patients with comorbidities in the first year after cirrhosis diagnosis, a finding corroborated by the synergy between the effects of cirrhosis and comorbidity. These findings suggest that treatment of comorbidities should be considered an integral part of clinical care for newly diagnosed cirrhosis patients. Successful treatment of comorbidity would reduce the mortality attributable to comorbidity itself, and also the mortality attributable to the synergy between comorbidity and cirrhosis.

The biological mechanisms of the cirrhosis-comorbidity synergy are unknown. We speculate that some comorbidities, such as congestive heart failure and renal disease, may reduce the body's ability to compensate for the circulatory consequences of a cirrhotic liver, and others may contraindicate treatments: chronic pulmonary disease may contraindicate β-blockers, heart failure may contraindicate insertion of a transjugular intrahepatic portosystemic shunt, and any severe comorbidity may contraindicate liver transplantation.13

The major strengths of our study were its size, its complete and long-term follow-up, and its 10 years of hospital-diagnosed comorbidities for each cirrhosis patient. The validity of the recorded diagnoses, including cirrhosis, is crucial for our findings. It appears that 15% of those we included may not have fulfilled diagnostic criteria for cirrhosis,29 but regardless of what is assumed about these patients' prognosis and comorbidity level, they could not have produced the dose-response relationship between comorbidity level and prognosis that we found if no such relationship existed.

We only included patients with a hospital diagnosis of alcoholic cirrhosis or unspecified cirrhosis. We did so because the ICD8 and ICD10 do not have separate codes for the histological stages of other chronic liver diseases, such as primary biliary cirrhosis and chronic viral hepatitis, of which cirrhosis is the end stage. For example, had we included patients at their first hospital diagnosis of primary biliary cirrhosis, we would have included many patients without cirrhosis. As it is, we fail to include some patients with nonalcoholic cirrhosis. However, alcoholism is the most common cause of cirrhosis in Denmark,7 and we showed that the association between comorbidity and mortality was essentially the same for cirrhosis patients with and without alcoholism; therefore, the association may be independent of the cause of cirrhosis.

The validity is not known for all the comorbidity diagnoses. Myocardial infarction, cancer, and diabetes have a very high validity,30 but even if a patient has several wrong or missing diagnoses, we may have classified him or her correctly with respect to the CCI categories that we used. Even if we misclassified some patients, the dose-response relationship between comorbidity level and prognosis indicates that few patients were misclassified.

We had no information on lifestyle factors other than alcohol abuse (for example, malnutrition and smoking), but they probably contributed to the higher comorbidity prevalence in cirrhosis patients than in the general population. They are also prognostic factors among cirrhosis patients4, 5 and could therefore have confounded our findings if they were more prevalent among patients with comorbidities. However, it is unlikely that malnutrition would completely explain our findings, because it should not be considered a confounder if it is caused by a comorbidity (for example, cancer).31

It was a limitation of our study that we did not have information on the cirrhosis stage at the time of diagnosis, but both the risk of cirrhosis-related and non–cirrhosis-related death increased with the comorbidity level. This pattern implies that many patients with severe comorbidity did not die from cirrhosis-related causes, so it is unlikely that they were in a more advanced cirrhosis stage at diagnosis than patients with less comorbidity. Also, patients who present at the hospital regularly for something other than cirrhosis may have a greater chance of having cirrhosis diagnosed in an early stage.

We hope that our findings contribute to both an understanding of the clinical course of cirrhosis and to clinical decision making.14, 15 The former goal might be achieved by studying the mechanisms of the cirrhosis-comorbidity synergy, the latter by including comorbidities in prognostic scoring systems for cirrhosis patients. To this end, it should be emphasized that the CCI was developed for unselected medical patients and would probably be even more strongly associated with prognosis if modified to be specific to cirrhosis patients.13 Another point to note is that comorbidity was not considered for inclusion in the Child-Pugh, Model for End-Stage Liver Disease, or Mayo systems,32–34 whereas it has long been recognized as an important prognostic factor in patients with upper gastrointestinal bleeding.35, 36 Comorbidity was therefore included in the widely used Rockall prognostic scoring system for upper gastrointestinal bleeding.37, 38

In conclusion, we have shown that comorbidity is strongly associated with mortality in patients with cirrhosis, and it appears that treatment of comorbidities should be a priority among newly diagnosed cirrhosis patients.

Ancillary