Potential conflict of interest: Nothing to report.
The clinical course of alcoholic cirrhosis, a condition with a high mortality, has not been well described. We examined prevalence, risk, chronology, and mortality associated with three complications of cirrhosis: ascites, variceal bleeding, and hepatic encephalopathy. We followed a population-based cohort of 466 Danish patients diagnosed with alcoholic cirrhosis in 1993–2005, starting from the date of hospital diagnosis and ending in August 2006. Data were extracted from medical charts during the follow-up period. Risk and mortality associated with complications were calculated using competing-risks methods. At diagnosis of alcoholic cirrhosis, 24% of patients had no complications, 55% had ascites alone, 6% had variceal bleeding alone, 4% had ascites and variceal bleeding, and 11% had hepatic encephalopathy. One-year mortality was 17% among patients with no initial complications, 20% following variceal bleeding alone, 29% following ascites alone, 49% following ascites and variceal bleeding (from the onset of the later of the two complications), and 64% following hepatic encephalopathy. Five-year mortality ranged from 58% to 85%. The risk of complications was about 25% after 1 year and 50% after 5 years for all patients without hepatic encephalopathy. The complications under study did not develop in any predictable sequence. Although patients initially without complications usually developed ascites first (12% within 1 year), many developed either variceal bleeding first (6% within 1 year) or hepatic encephalopathy first (4% within 1 year). Subsequent complications occurred in an unpredictable order among patients with ascites or variceal bleeding. Conclusion: Patients with alcoholic cirrhosis had a high prevalence of complications at the time of cirrhosis diagnosis. The presence and type of complications at diagnosis were predictors of mortality, but not of the risk of subsequent complications. (HEPATOLOGY 2010.)
We recently reported that each year 1 in 2,000 Danish citizens aged 45–64 years is diagnosed with alcoholic cirrhosis.1 Apart from their highly increased mortality,2, 3 little is known about their prognosis because the clinical course of alcoholic cirrhosis has not been systematically described.4 In this context, we define clinical course as the evolution of alcoholic cirrhosis after diagnosis.5
The prevalence of the classic cirrhosis complications at the time of diagnosis—ascites, variceal bleeding, and hepatic encephalopathy—and their association with mortality have previously been examined.3, 6–14 However, earlier studies were hospital- rather than population-based,3, 6–10, 15 small, comprising 100 or fewer patients,8, 9 or restricted to patients diagnosed before 1980,3, 6–8 when clinical management differed from recent practice. Most important, previous studies examined alcoholic and hepatitis-related cirrhosis together,11–14 despite evidence that prognosis depends strongly on etiology.2, 3 Thus, the risk of complications specifically among patients with alcoholic cirrhosis has not been investigated, and it remains unknown whether the clinical course of alcoholic cirrhosis is a predictable accumulation of complications culminating in death. This gap in knowledge makes it difficult to compare findings from clinical studies of cirrhosis patients with different complications, to assess whether treatments change the clinical course, and to plan interventions such as prophylactic treatments.
In this context, we studied all 466 residents of our hospital catchment area diagnosed with alcoholic cirrhosis during the 1993–2005 period, with follow-up extending to August 2006. We aimed to answer the following questions: What is the prevalence of complications at the time alcoholic cirrhosis is diagnosed? What is the mortality for patients before and after they develop complications? What are the risks and the typical chronological order of complications after cirrhosis diagnosis?
ICD, International Classification of Diseases; SNOMED, Systematized Nomenclature of Medicine; TIPS, transjugular intrahepatic portosystemic shunt.
Patients and Methods
This study was conducted in the hospital catchment area that includes the city of Aarhus, Denmark. The catchment area has a population of 365,000, and all Danish citizens have free, unlimited access to general practitioners and hospitals. In Denmark, in- or outpatient hospitalization is necessary to establish the diagnosis of cirrhosis, and the Aarhus catchment area's largest hospital, Aarhus University Hospital, has a specialist Department of Hepatology to which the other hospital departments refer patients with known or suspected cirrhosis. Nevertheless, we based our study on data from all hospitals in the catchment area to ensure that it was population-based, i.e., it included all patients within a geographic area.15
We used data from three healthcare registries to identify possible cirrhosis patients in the catchment area: the Danish National Patient Registry, which contains diagnoses made during all hospitalizations since 1977 and during all outpatient and emergency-room visits since 199516; the Danish Cause of Death Registry, which contains information supplied by clinicians and pathologists on up to four events or medical conditions that led to death for each decedent since 194317; and the Danish Pathology Registry, which stores diagnoses made at all histological examinations conducted since 1997. Using these registries, we identified 1,987 residents of our catchment area with a relevant diagnosis (see Appendix), and then reviewed their medical charts. We included in the study cohort all 466 patients who, according to the information recorded in their medical charts, had cirrhosis due in part or fully to alcohol abuse, were diagnosed between 1 January 1993 and 31 August 2005, and had not previously been examined for suspected cirrhosis. The remaining 1,521 patients did not fulfill all three inclusion criteria and were excluded.
The diagnosis of cirrhosis could rest on any combination of clinical, biochemical, imaging, hemodynamic, and liver biopsy findings, whereas the diagnosis of alcohol abuse was based on patients' and relatives' reports. Thus, patients were included in the cohort if the treating clinician was sufficiently confident of the diagnosis of alcoholic cirrhosis to record it in the medical chart. We did not negate clinicians' diagnoses of alcoholic cirrhosis on the basis of information that became available later, e.g., autopsy findings, nor did we include patients who were never believed to have cirrhosis until autopsy findings proved otherwise. The study inclusion date depended on the patient's presentation: For patients who presented with ascites, variceal bleeding, or hepatic encephalopathy and who had a history of alcohol abuse in the absence of another probable cirrhosis cause (viral hepatitis, autoimmune hepatitis, primary biliary cirrhosis, primary sclerosing cholangitis, alpha-1-antitrypsin deficiency, hemochromatosis, or Wilson's disease), it was usually the date of hospital admission. For patients without complications or with unclear cirrhosis etiology, it awaited clarification of the cirrhosis diagnosis and etiology.
From the medical charts we obtained the date on which patients presented with the following three complications: ascites, variceal bleeding, or hepatic encephalopathy. Ascites was defined as clinically detectable ascites, i.e., ascites seen only on ultrasound examination was excluded. Variceal bleeding was defined as clinically unequivocal bleeding from esophageal or gastric varices with hematemesis, a heart rate >100 beats per minute and a systolic blood pressure <100 mmHg, or a need for blood transfusion. Hepatic encephalopathy was defined as clinically overt hepatic encephalopathy, i.e., minimal hepatic encephalopathy18 was excluded. In practice, the diagnosis of hepatic encephalopathy was based on the patient's clinical presentation, usually supported by the blood ammonia level and/or a continuous reaction time test,19 and with differential diagnoses deemed unlikely.
Data on patients' alcohol consumption were extracted from the medical charts, as were data on liver transplantation, TIPS (transjugular intrahepatic portosystemic shunt) insertion, portosystemic shunt surgery, and spontaneous bacterial peritonitis, which was defined as >250 polymorphonuclear leukocytes per mm3 ascitic fluid.20 We recorded patients' current alcohol drinking status (abstinent or drinking) as reported when they were seen in the hospital and assumed that it remained unchanged until the next hospital contact. “Abstinence” was defined as complete abstinence or consumption of small amounts of alcohol on rare occasions, and “drinking” was defined as nonabstinence. Hence, our data collection took account of patients' changing drinking habits,21 and we used the data to categorize patients as abstainers (drinking during <5% of the follow-up time), intermittent drinkers (drinking during 5%–94% of the follow-up time), or persistent drinkers (drinking during ≥95% of the follow-up time).
Dates of death were obtained from the Danish Civil Registration System, which is continuously updated with dates of birth, death, and emigration.22 Causes of death were ascertained from the Danish Cause of Death Registry, with review of registry data to determine whether deaths were from cirrhosis or other causes. Death from liver failure, variceal bleeding, bacterial infection, or hepatocellular carcinoma counted as death from cirrhosis. Data linkage across data sources was made possible by the unique personal identifier issued to all Danish citizens at birth or immigration and used in all national databases and record systems.
At the time of inclusion into the study cohort, patients with alcoholic cirrhosis were classified into five categories according to the presence and type of cirrhosis complications: no complications; ascites alone; variceal bleeding alone; ascites and variceal bleeding; hepatic encephalopathy with or without ascites and variceal bleeding. Patients who developed complications during follow-up were reclassified into another category if appropriate. Based on our clinical experience and a previous Danish study,23 we assumed that a given complication carried the same prognosis whether it had been present at the time of cirrhosis diagnosis or developed later.
In all analyses follow-up ended at death or at censoring at the end of follow-up, on 31 August 2006. Analyses were conducted separately for the five complication categories and were based on the Aalen-Johansen estimator of the probability of having died or being in a particular category of cirrhosis complications at a particular point in time during follow-up.24, 25 Ninety-five percent confidence intervals were bootstrapped. We conducted three types of analyses which differed in the handling of complications during follow-up. In the first analysis complications were ignored, hence the Aalen-Johansen method was simplified to a Kaplan-Meier analysis yielding the cumulative mortality and the median survival time, i.e., the time to reach a cumulative mortality of 50%. In the second analysis complications were taken into account, but follow-up continued when they developed. On that basis we estimated the distribution of cirrhosis complication categories after 1 and 5 years of follow-up using the following categorization: alive without more complications; alive with more complications; dead without more complications; and dead with more complications. In the third analysis follow-up ended whenever complications developed, whereby the Aalen-Johansen method amounted to estimating the 1-and 5-year cumulative incidence (i.e., risk) of complications or death as competing events.25
We included 466 patients (71% men) at the time of cirrhosis diagnosis, and 449 (96%) were followed in the university Department of Hepatology. Of the remaining 17 patients, six had variceal bleeding and died in surgery before transfer to the Hepatology Department, three had severe comorbidity and were therefore transferred to other departments, two suffered sudden cardiac death before transfer to the Hepatology Department, and one was discharged from another hospital and referred to the Hepatology Department but died in the interim; it is not clear why the remaining five patients were not referred. Of the 466 patients, 30 (6%) were diagnosed on the basis of liver biopsy findings, the remainder on the basis of clinical, biochemical, imaging, and hemodynamic findings. At the time of inclusion the median age was 53 years (range: 27–84) and 79% of patients consumed alcohol. During a total observation time of 1,611 years, 299 (64%) patients died, six (1.3%) received a liver transplant, 26 (5.6%) had a TIPS inserted, and none had portosystemic shunt surgery. Thirty-six percent of patients maintained abstinence throughout follow-up, 43% were intermittent drinkers, and 21% were persistent drinkers. Patient characteristics and outcomes are shown in Table 1.
Table 1. Patient Characteristics and Outcomes for Each Category of Cirrhosis Complications
Ascites + Variceal Bleeding
Death from hepatocellular carcinoma counts as cirrhosis-related death.
Patients could be observed in more than one category, so the total number of observations (N=709) exceeds the number of patients included (N=466).
At inclusion, 114 (24%) patients had no complications, 254 (55%) had ascites alone, 29 (6%) had variceal bleeding alone, 20 (4%) had ascites and variceal bleeding, and 49 (11%) had hepatic encephalopathy alone or in combination with one or both of the other complications. The 114 complication-free patients were hospitalized for the following reasons at the time of diagnosis with alcoholic cirrhosis: gastrointestinal bleeding of nonvariceal origin (12%); hepatocellular carcinoma (6%); alcoholic hepatitis (4%); viral hepatitis (5%); other signs, symptoms, and/or blood chemistry suggestive of liver disease (53%); bacterial infection (11%); trauma (2%); or chronic disease not involving the liver (8%).
Mortality Before and After Complications
Patients Without Complications.
The median survival time for the 114 patients without initial complications was 48 months (Fig. 1). After 1 year, 68% were alive and complication-free, 15% were alive but had developed complications, 10% had died without developing complications, and 7% had died after developing complications. After 5 years, the corresponding proportions were 28%, 13%, 22%, and 35% (Table 2). Of the 24 deaths during follow-up, 18 (75%) were from cirrhosis, four (16%) from atherosclerotic disease, and two (8%) from oropharyngeal cancer (Table 1).
Table 2. Distribution of Cirrhosis Complication Categories At 1 (Top) and 5 Years (Bottom) After Entering the Five Complication Categories
Ascites + Variceal Bleeding
Hepatic Encephalopathy Alone or in Combination
Each cell shows the probability of being in a category and the associated 95% confidence interval. Please note that Fig. 1 shows the risk of death with or without having developed more complications (the total “Dead” group in this table), while Fig. 2 shows only the proportion who died without having developed more complications (the “Dead without more complications” group in this table).
Alive without more complications
Alive with more complications
Variceal bleeding alone
Ascites + variceal bleeding
Dead without more complications
Dead after developing more complications
Ascites + Variceal Bleeding
Hepatic Encephalopathy Alone or in Combination
Alive without more complications
Alive with more complications
Variceal bleeding alone
Ascites + variceal bleeding
Dead without more complications
Dead after developing more complications
Patients with Ascites.
At inclusion, 254 patients had ascites alone, and during follow-up a further 33 patients with no initial complications developed ascites (Table 1). Forty-eight patients (17%) had spontaneous bacterial peritonitis at first presentation with ascites. The median survival time for the 287 patients was 37 months from the onset of ascites (Fig. 1). After 1 year, 59% were alive and had not developed other complications, 12% were alive but had developed other complications, 15% had died without developing other complications, and 14% had died after developing other complications. After 5 years, the corresponding proportions were 32%, 9%, 25%, and 33% (Table 2).
Patients with Variceal Bleeding.
At inclusion, 29 patients had bleeding varices alone and 16 patients without initial complications had variceal bleeding during follow-up (Table 1). These 45 patients had a median survival time of 48 months from the onset of variceal bleeding (Fig. 1). During the first month after bleeding onset they had higher mortality than patients without complications (10% versus 4%), but long-term mortality was similar (Fig. 1). After 1 year, 64% of patients with variceal bleeding were alive without other complications, 16% were alive but had developed other complications, 11% had died without developing other complications, and 9% had died after developing other complications. After 5 years, the corresponding proportions were 27%, 8%, 18%, and 45% (Table 2). Seven of eight deaths in this patient category were from cirrhosis, and the one remaining death was from unknown causes (Table 1).
Patients with Both Ascites and Variceal Bleeding.
At inclusion, 20 patients had both ascites and variceal bleeding, and six of them (30%) had spontaneous bacterial peritonitis. During follow-up, 62 patients with a history of ascites developed variceal bleeding, whereas 12 patients with a history of variceal bleeding developed ascites (Table 1). The median survival time for the total 94 patients was 13 months from the onset of the later of the two complications (Fig. 1). After 1 year, 47% were alive without hepatic encephalopathy, 4% were alive but had developed hepatic encephalopathy, 31% had died without hepatic encephalopathy, and 18% had died after developing hepatic encephalopathy. After 5 years the corresponding proportions were 17%, 4%, 44%, and 36% (Table 2).
Patients with Hepatic Encephalopathy.
At inclusion, 49 patients had hepatic encephalopathy, and during follow-up hepatic encephalopathy developed in nine patients who never had complications, in 66 patients with a history of ascites alone, in 10 patients with a history of variceal bleeding alone, and in 35 with a history of both ascites and variceal bleeding (Table 1). Eighty-five patients had ascites when they first developed hepatic encephalopathy, and 26% of these had spontaneous bacterial peritonitis. The 169 patients with hepatic encephalopathy had a median survival time of 2.4 months from its onset; 45% died within 1 month, 64% died within 1 year, and 85% died within 5 years (Fig. 1, Table 2).
Risk and Chronology of Complications
Ascites was the most frequent first complication (12% of patients developed this complication within the first year after cirrhosis diagnosis), but nearly as many patients developed either variceal bleeding (6%) or hepatic encephalopathy (4%) as their first complication. Hence, 22% of patients developed one of the three complications under study during the first year after being diagnosed with alcoholic cirrhosis (Fig. 2).
Patients with ascites were equally likely to develop variceal bleeding or hepatic encephalopathy as their next complication (1-year risk = 12% and 15%, respectively). Similarly, patients with variceal bleeding were equally likely to develop ascites or hepatic encephalopathy next (1-year risk = 13% and 11%, respectively) (Fig. 2). Thus, there was no evidence of a particular chronological pattern in the order of developing the three complications. Despite the lack of a predictable sequence of clinical events, all patients without hepatic encephalopathy had about a 25% risk of developing a complication within 1 year after entry into a given complication category, and a 50% risk of developing one within 5 years (Fig. 2).
We found that Danish patients with alcoholic cirrhosis most frequently presented either with ascites alone (55%) or without complications (24%). The presence and type of complications were strong predictors of 1-year mortality, which ranged from 17% for patients without complications to 85% for patients with hepatic encephalopathy. The risk of new complications was similar for all patients. Complications did not develop in any particular chronological sequence, suggesting that the clinical course of alcoholic cirrhosis is a random series of clinical events and not a predictable accumulation of complications.
Our risk estimates were based on complete follow-up of a population-based patient sample under specialist care and thus are likely to generalize to alcoholic cirrhosis patients in other settings, particularly given that treatment guidelines in Western countries were already established at the beginning of our study period and follow similar principles. However, the risk estimates could be biased if some patients did not have cirrhosis or did not have the complications. This is unlikely, based on a study of Danish alcohol-abusing men which showed that clinical cirrhosis diagnoses have a high positive predictive value when validated against biopsy findings.26 Moreover, it is highly plausible that all first-time episodes of evident ascites, variceal bleeding with hematemesis, and overt hepatic encephalopathy were recorded in the medical charts.
The 55% prevalence of ascites at cirrhosis diagnosis among patients in our study exceeds the 30%–40% prevalence estimated in most previous studies, whereas the prevalence of the other two complications is similar to that reported in those studies.3, 6, 7, 10 The reasons for the higher prevalence of ascites in our study are unclear. A still higher prevalence was found in a Norwegian study based on 100 patients with alcoholic cirrhosis seen in a single medical department in 1984–1988.9 That study reported a 67% prevalence of ascites and a 34% prevalence of variceal bleeding at the time of cirrhosis diagnosis, and those high prevalences might be explained by selective inclusion of patients who were referred to a specialist department because of advanced disease. Our study, by contrast, was population-based and therefore not vulnerable to such an inclusion bias.
Mortality among complication-free patients in our study was higher than that reported in the two previous studies of alcoholic cirrhosis patients without complications,3, 7 despite those studies' inclusion of patients diagnosed between 1951 and 1976, when treatment strategies were less aggressive. A possible explanation for this apparent lack of improvement in clinical management of cirrhosis is the 47% prevalence among our patients of comorbidities or complications other than those we considered in our analysis. Comorbidity has recently been demonstrated to increase both all-cause and cirrhosis-related mortality,27 and its importance is corroborated by the observation that a quarter of our patients did not die from cirrhosis, compared with 15%–20% in the older studies.3, 7 Differences in alcohol consumption also may be of importance; the proportion of abstainers in our cohort matched that in the older studies, but in those studies only complete teetotalers counted as abstainers.3, 7
Among patients in our study, mortality increased further following the development of complications, in accordance with the existing literature.28 Probably the higher proportion of persistent drinkers among patients with complications contributed to this association. Mortality among patients with variceal bleeding has previously been found to be similar in those with and without a history of ascites,28 but our results and those from a recent German study demonstrate that this is not the case.29 A likely explanation for the emerging importance of ascites among patients with variceal bleeding is that bleeding is less fatal now than it was in the past.30 In fact, the mortality of patients with complications was consistently lower in our study than in older studies.3, 6, 7, 10, 31
The largest earlier study, including 122 Spanish patients with alcoholic cirrhosis and 171 patients with nonalcoholic cirrhosis,11 reported that the risk of developing ascites, variceal bleeding, or hepatic encephalopathy increased steadily by 7%–10% per year in the cohort as a whole.11–14 This is consistent with our finding that 49% of patients without complications at cirrhosis diagnosis developed complications within 5 years. At the same time, the risk in our study was much higher during the first year (22%) than during the following 4 years (27%, or about 7% per year). In the Spanish study, patients were not included when the clinical diagnosis was made, but when it had been confirmed by a liver biopsy in a specialist unit.11 However, patients at highest risk of complications may not have survived from clinical diagnosis to inclusion, and the risk of complications could therefore have been underestimated. Furthermore, although our study corroborates previous findings that ascites is usually the first complication to appear,11, 28 we also found a high risk of variceal bleeding or hepatic encephalopathy as the first complication. This indicates that patients with alcoholic liver cirrhosis should always be considered at risk of all three complications. Finally, we found that ascites formation was associated with a doubled risk of variceal bleeding within the next year, from 6% to 12%, and if patients with ascites developed variceal bleeding, their life expectancy dropped to 13 months. Therefore, patients with a history of both ascites and variceal bleeding should be considered for liver transplantation, as should patients with hepatic encephalopathy.
There have been attempts to describe the clinical course of cirrhosis as a progression through successive stages defined by the presence of particular complications. Our study shows that such a staging system cannot be based on ascites, variceal bleeding, and hepatic encephalopathy because these complications do not develop chronologically. The same conclusion applies to nonbleeding varices, bleeding varices, and ascites, which have been studied by D'Amico et al.28 Hepatocellular carcinoma, alcoholic hepatitis, and bacterial infection also may occur at any time during the clinical course of cirrhosis, and so it appears that cirrhosis complications develop in a random sequence. Therefore, a staging system may need to be based on factors that determine the development of complications, such as metabolic liver function and portal pressure,28 but alcohol consumption could also be important, as indicated by its strong association with mortality.3 However, this study was not designed to examine whether particular patient characteristics accelerate the clinical course or predispose patients to developing one particular complication instead of another, and we could not reach a firm conclusion on the prognostic role of alcohol consumption without also considering the possibly confounding effects of factors such as gender, age, comorbidity, treatment, and compliance. Such an analysis was not possible within the framework of the present study.
In conclusion, we systematically examined the clinical course of alcoholic cirrhosis among patients treated in a Danish geographic region. Our findings demonstrate that patients with alcoholic cirrhosis have a high prevalence of complications at the time of diagnosis, and that these complications are strong predictors of 1-year mortality, but not of the risk of developing more complications. In addition, because complications develop in a seemingly random sequence, the clinical course of alcoholic cirrhosis cannot be determined based on the presence or absence of particular cirrhosis complications.
We retrieved the medical charts of all patients in our hospital catchment area who in the period from January 1, 1993 through August 31, 2005 had a liver biopsy showing cirrhosis (defined as a SNOMED [Systematized Nomenclature of Medicine] code of M495xx or M496xx), or one or more of the diagnoses listed recorded in the National Patient Registry or in the Cause of Death Registry. Diagnoses were coded using the 8th edition of the International Classification of Diseases (ICD-8) through 1993 and the 10th edition (ICD-10) thereafter.