Long-term prognosis for transplant-free survivors of paracetamol-induced acute liver failure

Authors


Dr P. Jepsen, Department of Medicine V (Hepatology and Gastroenterology), Aarhus University Hospital, Nørrebrogade, 8000 Aarhus C, Denmark.
E-mail: pj@dce.au.dk

Abstract

Aliment Pharmacol Ther 2010; 32: 894–900

Summary

Background  The prognosis for transplant-free survivors of paracetamol-induced acute liver failure remains unknown.

Aim  To examine whether paracetamol-induced acute liver failure increases long-term mortality.

Methods  We followed up all transplant-free survivors of paracetamol-induced acute liver injury, hospitalized in a Danish national referral centre during 1984–2004. We compared age-specific mortality rates from 1 year post-discharge through 2008 between those in whom the liver injury led to an acute liver failure and those in whom it did not.

Results  We included 641 patients. On average, age-specific mortality rates were slightly higher for the 101 patients whose paracetamol-induced liver injury had caused an acute liver failure (adjusted mortality rate ratio = 1.70, 95% CI 1.02–2.85), but the association was age-dependent, and no survivors of acute liver failure died of liver disease, whereas suicides were frequent in both groups. These observations speak against long-term effects of acute liver failure. More likely, the elevated mortality rate ratio resulted from incomplete adjustment for the greater prevalence of substance abuse among survivors of acute liver failure.

Conclusions  Paracetamol-induced acute liver failure did not affect long-term mortality. Clinical follow-up may be justified by the cause of the liver failure, but not by the liver failure itself.

Introduction

Acute liver failure is an immediate life-threatening condition that in Denmark and many other countries is caused primarily by paracetamol poisoning.1–3 The treatment strategy – securing patient survival until effective liver regeneration takes over and resorting to liver transplantation if this strategy fails – relies on the assumption that the liver is capable of complete regeneration leading to clinical recovery.4–6 The purpose of our study was to evaluate this assumption by studying long-term mortality among survivors of paracetamol-induced acute liver failure.

A recent case series suggested that nonparacetamol drug-induced liver injury with jaundice could play a role in the long-term risk of chronic liver disease, but without a comparison group, a firm conclusion could not be reached.7 The choice of comparison group is critical because it is difficult to disentangle the long-term effects of the acute liver failure from the confounding effects of the other characteristics of patients who develop acute liver failure. Hence, we restricted our study to patients at risk of acute liver failure, viz. patients with severe paracetamol-induced liver injury. Among them, we compared post-discharge mortality between those who had or had not experienced an acute liver failure. To corroborate the necessity of using restriction to reduce confounding, we also compared the mortality of patients with paracetamol-induced liver injury without acute liver failure with the mortality in the general population.

Materials and methods

Study population

According to national guidelines, Danish citizens hospitalized with severe paracetamol-induced liver injury [i.e. with the International Normalized Ratio (INR) above 1.7] must be considered for transfer to The Liver Unit at Rigshospitalet, Copenhagen, Denmark’s only liver transplantation centre. The decision to transfer patients is made by specialists at The Liver Unit on an individual basis. We identified all patients who were hospitalized for paracetamol-induced liver injury in The Liver Unit during 1984–2004, had an INR above 1.7, but survived without a liver transplant and were still alive 1 year after discharge. We had no data on other patients. Among the 656 identified patients, we excluded 15 who at the time of the paracetamol-induced liver injury had chronic liver disease, identified through the Danish National Patient Registry, which has recorded hospital contacts in Denmark since 1977.8 The registry data include admission and discharge dates and discharge diagnoses coded by physicians using the 8th revision of the International Classification of Diseases (ICD-8) through 1993 and the 10th revision (ICD-10) thereafter. Chronic liver disease was defined by a history of a diagnosis code for alcoholic cirrhosis (ICD-8: 571.09; ICD-10: K70.3), biliary cirrhosis (ICD-8: 571.90–91; ICD-10: K74.3-5), chronic hepatitis (ICD-8: 571.93, 573.02; ICD-10: B18.x, K73.x), or unspecific cirrhosis (ICD-8: 571.92, 571.99; ICD-10: K74.0-2, K74.6). Simultaneous overdose of drugs other than paracetamol was not a criterion for inclusion or exclusion.

Patient data

We used data from the patients’ medical charts to ascertain the diagnosis of acute liver failure, defined by coagulopathy, jaundice and hepatic encephalopathy grade 2–4.9 We also obtained information on the presence or absence of alcohol abuse; other substance abuse; schizophrenia or use of antipsychotic medication; affective disorder or use of antidepressants; borderline personality disorder; and eating disorder. We supplemented the chart data with discharge diagnoses of the same conditions in the National Patient Registry: alcohol abuse (ICD-8: 291.xx, 303.xx except 303.90; ICD-10: F10.2-9), other substance abuse (ICD-8: 294.30, 294.38, 304.xx; ICD-10: F11.x to F19.x), schizophrenia (ICD-8: 295.xx; ICD-10: F2x.x), affective disorder (ICD-8: 296.xx, 298.xx; ICD-10: F3x.x), borderline personality disorder (ICD-8: none; ICD-10: F60.3) and eating disorder (ICD-8: 294.19; ICD-10: F50.x). If these conditions were noted in a patient’s medical chart at admission for paracetamol-induced liver injury, they were considered present throughout the follow-up. If they were identified through the National Patient Registry, they were considered present from the date of the first hospital discharge with the diagnosis.

We also obtained information on the following conditions, using only discharge diagnoses in the National Patient Registry: cancer (ICD-8: 140.xx to 195.xx; ICD-10: Cxx.x), chronic obstructive lung disease (ICD-8: 491.xx, 492.xx; ICD-10: J42.x, J43.x, J44.x), diabetes (ICD-8: 249.xx, 250.xx; ICD-10: E10.x to E14.x), arterial hypertension (ICD-8: 40x.xx; ICD-10: I10.x to I15.x) and ischaemic heart disease (ICD-8: 41x.xx; ICD-10: I20.x to I25.x). These conditions were considered present from the date of the first hospital discharge with the diagnosis.

Mortality data

Dates of death were obtained from the Central Office of Civil Registration, where data are updated continuously with no loss to follow-up.10 Causes of death were ascertained from the Cause of Death Registry, which contains information on up to four events or medical conditions that led to death for each decedent. The data are supplied by the physician responsible for the patient’s treatment at the time of death.11 We reviewed the registered data to define a single cause of death for each patient; the investigators were blinded to patients’ acute liver failure status.

Statistical analyses

Follow-up began 1 year after the date of discharge following paracetamol-induced liver injury and ended at death or on 31 December 2008, whichever came first. The 1-year delay in follow-up was introduced to avoid bias due to early deaths related to the poisoning or its psychiatric cause (e.g. risk of completed suicide is particularly high up to 1 year after a paracetamol poisoning).12 We used the Mann–Whitney test for the continuous variables, i.e. patient age and calendar time, and Pearson’s chi-square test for the other variables to test the hypotheses that patients with or without acute liver failure had similar characteristics at inclusion. We used a P-value below 0.05 to denote statistically significant nonsimilarity.

We compared the mortality rate ratio (specifically, the hazard ratio) for patients with and without acute liver failure using Cox proportional hazards regression, and we adjusted for the effects of gender, calendar year, alcohol abuse, other substance abuse, schizophrenia or use of antipsychotics, affective disorder or use of antidepressants, borderline personality disorder, eating disorder, cancer, chronic obstructive lung disease, diabetes, arterial hypertension, and ischaemic heart disease. As we followed up patients from a time point defined by recovery from acute paracetamol-induced liver injury, we anticipated that patient age was a stronger predictor of death than time since hospital discharge, and therefore a more relevant time scale for our analyses.13 Using natural splines, we plotted the age-specific adjusted mortality rate ratios for acute liver failure.14

Comparison with population mortality

Through the Central Office of Civil Registration, we identified a control group from the general population, the aim being to substantiate the excess mortality for patients hospitalized for paracetamol-induced liver injury. We included 100 gender-, age-, and calendar time-matched Danish citizens per patient without acute liver failure, and they were drawn at random using risk set sampling.15 Using Cox regression and natural splines, as described above, we plotted the age-specific mortality rate ratios for the general population sample vs. patients with paracetamol-induced liver injury without liver failure.

Results

We included 641 transplant-free survivors of paracetamol-induced liver injury, and 101 (16%) of these had acute liver failure. Among the patients with acute liver failure, 16% had hepatic encephalopathy of grade 2; 51% of grade 3; and 33% of grade 4. The median follow-up time was 7.6 years, the maximum 23.2 years; 86 patients (13%) died during the follow-up, including 23 patients in the acute liver failure group.

Patients with acute liver failure were more likely than patients without acute liver failure to be female, and they were older, and more likely to abuse alcohol and other substances (Table 1). Also the age-specific prevalence of substance abuse was higher among patients with a history of acute liver failure than among patients without such history. For example, at age 45 years, the prevalence of alcohol abuse was 58% in the acute liver failure group (14 of 24 patients under observation) compared with 29% in the group with milder liver injury (29 of 101 patients under observation) (Table 2).

Table 1.   Characteristics of patients at inclusion. For the continuous variables, age and year of inclusion, the numbers shown are the median and 25th and 75th percentiles
 Acute liver failureNo acute liver failureP-value
Age43 (33–55)35 (24–47)<0.001
Women73 (72%)331 (61%)0.05
Year of inclusion1999 (1996–2003)2001 (1996–2003)0.50
Substance abuse
 Alcohol abuse47 (47%)145 (27%)<0.001
 Other substance abuse14 (14%)34 (6.3%)0.01
Psychiatric disease
 Schizophrenia or use of antipsychotics16 (16%)59 (11%)0.21
 Affective disorder or use of antidepressants42 (42%)207 (38%)0.61
 Borderline personality disorder2 (2.0%)12 (2.2%)0.83
 Eating disorder08 (1.5%)0.46
Somatic disease
 Cancer1 (1.0%)4 (0.74%)0.72
 Chronic obstructive lung disease1 (1.0%)6 (1.1%)0.68
 Diabetes3 (3.0%)15 (2.8%)0.83
 Arterial hypertension4 (4.0%)12 (2.2%)0.50
 Ischaemic heart disease2 (2.0%)10 (1.9%)0.75
Total101 (100%)540 (100%) 
Table 2.   Characteristics of the 641 patients with paracetamol-induced liver injury, by current age. Columns show the number of patients in the cohort who are observed at a particular age and their characteristics. Patients were followed up from 1 year after discharge after a paracetamol-induced liver injury, so a patient who was 33 years old at discharge and died at age 49 is under observation from age 34 to age 49
 Current age (years)
20304045506070
  1. * Schizophrenia including use of antipsychotics, affective disorder including use of antidepressants, borderline personality disorder, or eating disorder.

  2. † Cancer, chronic obstructive lung disease, diabetes, arterial hypertension, or ischaemic heart disease.

Number of patients under observation
 Acute liver failure811202419202
 No acute liver failure76113112101825524
Women
 Acute liver failure (%)7 (88)7 (64)15 (75)17 (71)15 (79)15 (75)2 (100)
 No acute liver failure (%)68 (89)63 (56)59 (53)62 (61)48 (59)35 (64)16 (67)
Alcohol abuse
 Acute liver failure (%)02 (18)11 (55)14 (58)12 (63)11 (55)2 (100)
 No acute liver failure (%)3 (4)10 (9)24 (21)29 (29)34 (41)22 (40)13 (54)
Other substance abuse
 Acute liver failure (%)04 (36)2 (10)3 (12)3 (16)2 (10)1 (50)
 No acute liver failure (%)1 (1)5 (4)8 (7)6 (6)7 (9)4 (7)3 (12)
Psychiatric disease*
 Acute liver failure (%)2 (25)4 (36)7 (35)9 (38)8 (42)11 (55)2 (100)
 No acute liver failure (%)15 (20)36 (32)39 (35)43 (43)37 (45)31 (56)16 (67)
Somatic disease†
 Acute liver failure (%)002 (10)3 (12)5 (26)5 (25)1 (50)
 No acute liver failure (%)01 (1)4 (4)12 (12)10 (12)18 (33)6 (25)

On average, age-specific mortality rates were higher for patients with a history of acute liver failure (adjusted mortality rate ratio = 1.70, 95% CI 1.02–2.85). However, the mortality rate ratio was strongly age-dependent, so that 25-, 45-, or 75-year-old people who had more than 1 year previously been hospitalized with paracetamol-induced acute liver failure had higher mortality than people of equal age who had more than 1 year previously been hospitalized with paracetamol-induced acute liver injury without liver failure, and vice versa among 35- or 60-year-old people (Figure 1). However, the earlier temporary failure of liver function was not the cause of the excess mortality: no patient with a history of acute liver failure was recorded to die from liver disease (data on causes of death were unavailable for three deceased). In contrast, six patients without acute liver failure died of acute or chronic liver disease. Suicides were very frequent in both groups (26 suicides = 37% of the 76 deaths with data on causes of death), as were accidents, cancer and cardiovascular disease (Table 3).

Figure 1.

 Mortality rates for patients with a history of paracetamol-induced acute liver failure (black) and for a random population sample (grey) compared with that for patients with a history of paracetamol-induced liver injury without liver failure (the dotted horizontal line). The thin lines show 95% confidence limits. Hence, the black line shows that at age 45–50 years, primarily, mortality was higher for patients with a history of paracetamol-induced acute liver failure than for patients with a history of paracetamol-induced liver injury without acute liver failure. The grey line shows that mortality before age 60 years was much lower for the general population sample than for patients with a history of paracetamol-induced liver injury without acute liver failure. The horizontal boxplots illustrate the age distribution (median, quartiles, and extremes) at the time of paracetamol-induced liver injury with or without acute liver failure (ALF), and the superimposed black dots mark the ages at which deaths occurred.

Table 3.   Causes of death. A total of 86 patients died during follow-up, but this Table excludes the three patients with acute liver failure and seven patients without acute liver failure with no available data on cause of death
Cause of deathAcute liver failureNo acute liver failure
Acute liver disease01 (2%)
Chronic liver disease05 (9%)
Alcoholism05 (9%)
Cancer4 (20%)6 (11%)
Cardiovascular disease3 (15%)7 (13%)
Diabetes1 (5%)0
Pneumonia02 (4%)
Respiratory disease2 (10%)0
Gastrointestinal bleeding01 (2%)
Dementia01 (2%)
Epilepsy01 (2%)
Accident2 (10%)7 (13%)
Suicide8 (40%)20 (36%)
Total20 (100%)56 (100%)

Mortality before age 60 years was clearly higher for patients with a history of paracetamol-induced liver injury without acute liver failure than for the general population sample (Figure 1). Hence, the mortality in the general population was not a relevant estimate of the expected mortality among the 91% of our patients who were younger than 60 years at discharge after paracetamol-induced liver injury.

Discussion

We followed up transplant-free survivors of paracetamol-induced liver injury and found that those who had suffered an acute liver failure had marginally higher mortality, however, for causes not attributable to their earlier liver failure. We also showed that survivors of paracetamol-induced liver injury who were younger than 60 years of age had a markedly higher mortality than the general population, largely due to suicides and accidents.

It is a long-standing and basic assumption that there is complete clinical recovery after a reversible loss of liver function. This assumption is rooted in the well-known regenerative capacity of the liver, but it has not previously been challenged by a long-term clinical follow-up study. Our data support the classical assumption. We found a marginal increase in mortality after acute liver failure, but it was age-dependent and not due to liver-related deaths. Rather, we expect the association to be an overestimate due to incomplete control of confounding by substance abuse, which was only crudely measured as absent or present.

The validity of the clinical diagnoses of acute liver failure is crucial because invalid diagnoses would have led us to underestimate the true effect of acute liver failure.16 However, we believe that the diagnoses were valid because they were made by specialists in a liver transplantation centre and required clinically overt hepatic encephalopathy, i.e. grade 2–4.

The findings in our study accord with a recent follow-up study in which patients with and without hepatotoxic paracetamol overdose had parallel survival curves from 1 year after hospital discharge.17 In that study, however, at most only 15 deaths were observed more than 1 year after hospital discharge. Likewise, a study of postdischarge liver biopsies from patients with paracetamol-induced liver injury supplied histological evidence of reversibility,6 and complete regeneration has also been described after surgical liver injury in animals and humans4, 5 and after severe toxic liver injury in rats.18

Our comparison with mortality in the general population is consistent with our previous findings that, for several years after hospitalization for paracetamol poisoning, the risks of being diagnosed with a psychiatric disease and/or completing suicide are greatly elevated.12, 19 Our analyses also demonstrate that a general population sample is not the proper comparison group for evaluating long-term effects of acute liver injury: those pre-existing patient characteristics that lead to experiencing an acute liver injury heavily influence prognosis. Therefore, we conducted the study within a cohort of patients with a history of paracetamol-induced liver injury and excluded those who died within a year postdischarge. These design features minimized confounding from patient characteristics such as suicidal intent as opposed to therapeutic misadventure, because we can expect similar characteristics – 82% with suicidal intent – in equal-age Danish patients with or without acute liver failure after paracetamol poisoning.20

In conclusion, paracetamol-induced acute liver failure did not increase long-term mortality, but many patients hospitalized for paracetamol-induced liver injury eventually completed suicide. Hence, our findings reiterate that all patients hospitalized for paracetamol-induced liver injury should undergo psychiatric evaluation and follow-up, whereas follow-up by hepatologists is not indicated.

Acknowledgement

Declaration of personal and funding interests: None.

Ancillary