Article Text

Download PDFPDF

Mortality from cirrhosis: lack of progress over the last 35 years
  1. C Gluud
  1. Correspondence to:
    Dr C Gluud
    Copenhagen Trial Unit, Centre for Clinical Intervention Research, Department 7102, Rigshospitalet, Copenhagen University Hospital, DK 2100 Copenhagen, Denmark;

Statistics from

Liver cirrhosis mortality does not seem to have changed significantly during the last 30 years in the Oxford region! How come? There are risks of erroneous findings (patient mix consequences, wrong sampling unit, diagnostic misclassification, type II errors), lack of beneficial effects of interventions considered to be effective, and incorrect implementation of interventions

The findings of Roberts and colleagues1 in this issue of Gut show a lack of improvement in liver cirrhosis mortality during the years 1968 to 1999 (see page 1615). This must be provoking, disappointing, and sobering reading for many.

Roberts and colleagues1 based their findings on 8192 patients admitted to hospitals in the Oxford region of Southern England. The analyses were based on discharge and death certificate statistics. One year after admittance, the mortality rate was 34%, and remained so during the entire 30 year observation period. One year after admittance, the standardised mortality rate was 16.3 times that of the general population. These findings were robust to different analytical strategies. The disappointing results during the 30 year period were obtained both for the total group of 8192 patient admissions as well as for the diagnostic subgroups individually. The data show how deadly liver cirrhosis has been and continues to be.

Why does cirrhosis mortality remain unaffected despite our impression that it is decreasing? We expect decreasing mortality due to the use of many new interventions. Interventions such as liver transplantation,2 endoscopic treatment for bleeding varices,3,4 prevention of bleeding and rebleeding from varices with beta blockers,3,4 antiviral drugs for hepatitis B or C,5,6 and antibiotics for preventing and treating complications to cirrhosis,3,7–9 have all been introduced during the study period.

Three major causes may explain the contrast between the finding of no significant change in mortality on the one hand and our expectation of decreasing mortality on the other: erroneous findings, misguided perceptions about the effectiveness of interventions, or incorrect implementation of interventions.


A change in “patient mix” over time is a plausible possibility for erroneous findings. Nurses and physicians feel that the people they care for and treat are becoming more severely ill. The workload connected with the individual patient is rising, and so does the number of hospital admissions. Outpatient clinics are growing at a rate of approximately 2% per year. Part of this growth could be due to handling less diseased patients as outpatients. Accordingly, more severely diseased patients are admitted to hospital wards. This would raise mortality. But improved treatments keep the mortality at a constant level. Therefore, no net change in mortality is observed.

The patient mix could be influenced by diagnostic drift. Thirty years ago the diagnoses of “non-alcoholic steatohepatits” and “non-alcoholic fibrosis and cirrhosis” were rarely used. These diagnoses have become trendy over the years.10 This may have moved the less severely diseased patients from the alcoholic cirrhosis group to the “new” diagnostic entities. Studies on the “new” non-alcoholic liver diagnoses rarely exclude alcohol problems by adequate means,11–14 so why should clinical practice?

Roberts and colleagues1 may have chosen the “wrong” sampling unit? During the past decades population based liver cirrhosis mortality rates have increased significantly in England, in contrast with other EU countries where it is falling.15 The increase in alcohol consumption may lead to increased mortality rates. This could impact on the patient mix and the findings. However, the findings of Roberts and colleagues1 concur with data from other regions of England16 and from Denmark.17

Roberts and colleagues1 used administrative discharge data for their identification of cases. Diagnostic discharge data are not entirely accurate and liver cirrhosis may have been misclassified in 5–10% of cases, according to the Danish experience.18 Such misclassification however would not explain the lack of improvement in liver cirrhosis mortality unless one could point to a systematic change connecting misclassification over time with changed mortality.

We must consider the possibility of a type II error—no improvement was detected although it actually happened. Nobody can exclude this possibility. The type II error is in fact substantial. Mortality during the period 1994 to 1999 had an odds ratio of 0.90 (95% confidence interval (CI) 0.74–1.09) compared with mortality in the period 1968–1973 (table 4 in Roberts and colleagues1). This corresponds to the finding that mortality may have decreased by 26% or increased by 9%. Only much larger studies, preferably including national or international data, may resolve this issue.


The second major cause for the contrast between no significant improvement in mortality on the one hand and our expectation of improvements in mortality due to new methods for prevention, diagnosis, treatment, and care on the other is misguided expectations. Could it be that there has in reality been no improvement in mortality during the last 30 years?1

We have no randomised clinical trials demonstrating that cirrhosis mortality can be significantly decreased by liver transplantation. However, we have cohort studies and matched controlled and simulated studies suggesting that this may be so, but these analyses only point to the fact that Child class C patients may have a gain in survival.2,19 Furthermore, less than 25 Child class C patients undergo liver transplanted per year in the Oxford region. This number could not influence the statistics for a whole region.

Endoscopic banding in the primary prevention of variceal bleeding may prevent bleeding but endoscopic banding does not seem to have significant effects on mortality.20 Beta blockers are also recommended for primary prevention3 but we have been unable to demonstrate any significant effect of beta blockers on mortality in high quality trials (Chen W et al, unpublished observations).

Terlipressin and endoscopic interventions may significantly decrease mortality in patients with acute variceal bleeding. But we still lack information on the influence of these interventions on long term mortality.3,21,22

Prevention of rebleeding from varices with beta blockers, endoscopic measures, or transjugular intrahepatic portosystemic stent shunt (TIPS) may have significant effects on mortality but the results are modest.3,23–25 TIPS may remove more ascites but cause more hepatic encephalopathy and does not seem to have significant effects on mortality.26

Much marketing has gone into interventions for chronic hepatitis B and chronic hepatitis C. We do not have any firm knowledge on how interferons affect mortality in patients with chronic hepatitis B.27 Lamivudine seems to affect the clinical course in patients with chronic hepatitis B and advanced liver disease.28 However, we have no convincing evidence showing a significant effect of lamivudine28 on the mortality of chronic hepatitis B patients. The situation for patients with chronic hepatitis C is not much better. Combination therapy with interferon and ribavirin seems to be able to reduce the composite outcome measure of cirrhosis, hepatocellular carcinoma, and death.29 Combination therapy significantly reduced the composite outcome measure in all 9991 patients entered into 72 randomised clinial trials (relative risk 0.49 (95% CI 0.25–0.96; number needed to treat to benefit one patient 419 (95% CI 290–4523)) but not in naive, relapsers, or non-responders individually.29 Furthermore, we found no significant effect of combination therapy on mortality as an isolated outcome.29

Antibiotics are recommended for preventing and treating complications of cirrhosis.3,7–9 Although antibiotics seem to reduce short term mortality we urgently need data on their influence on long term mortality.


If the explanation that the interventions do not work is unpalatable, then try this: the interventions worked only a little in clinical research but were incorrectly applied in clinical practice.

We recently assessed the degree of concurrence between research evidence and clinical practice in Denmark.30,31 We evaluated the agreement between research evidence in Cochrane Hepato-Biliary Group systematic reviews32 and reported on the use of clinical interventions for typical patients presenting no contraindications. The Cochrane Hepato-Biliary Group had completed 28 systematic reviews on 36 interventions that were available in Denmark in 2002. Based on the reviews, three interventions with significant beneficial effects on clinical outcomes were classified as “evidence based” (n-acetylcysteine for paracetamol overdose,33 terlipressin for bleeding oesophageal varices,21 and antibiotics for patients with cirrhosis and gastrointestinal bleeding7) whereas 19 were classified as “possibly evidence based” and 14 as “not evidence based”. Questionnaires on reported use and perceived intervention effects were mailed to senior physicians practising in Danish hospitals and being responsible for the treatment of liver patients. The proportion of physicians who reported that they never used the three evidence based interventions varied considerably (2%, 57%, and 62%, respectively). The perceived intervention effect, duration of clinical experience, employment as head of department, and university hospital employment were significant predictors of more frequent use of evidence based interventions. Physicians also reported that they used the interventions that were not evidence based more often if they were employed at a university hospital.

So we observed considerable disagreements between research evidence and reported clinical use in Denmark. Although I have no data to prove it, similar or other mechanisms in Oxford could explain incorrect implementation of interventions.


As always, when things do not add up more than one thing may be wrong. Could it be a combination of erroneous findings, misguided expectations, and incorrect implementation? The findings in the study of Roberts and colleagues1 as well as the increased liver cirrhosis mortality rates in England over time—in contrast with other countries—call for further epidemiological studies in the UK and abroad. In addition to one year mortality, these studies should also examine five and 10 year mortality changes over time.

We may also need to adjust our expectations to the intervention effects. May be the interventions we use are not as effective as we imagine. We therefore urgently need to develop more effective interventions. In addition to finding new interventions for alcoholic liver cirrhosis, for example,34–37 we must develop preventive measures for the deadly consequences of the increasing alcohol problems in our societies. We also need to scrutinise the beneficial and harmful effects of the intervention we employ in daily clinical practice. In order to accomplish these developments we must become much better in collaborating. Through collaboration we can conduct much larger randomised clinical trials carried out with much better methodology.38–43

We also need more public support for clinical research and possibly less industry involvement.44,45 Finally, we need full transparency in clinical research with publication of and public access to all protocols, protocol amendments, and clinical data.46

Finally, all hospitals and departments should regularly check their clinical practices and clinical guidelines. We need collaboration on getting research into clinical practice. All departments need a team of systematic reviewers who are able to conduct high quality systematic reviews.47 Only by being able to do systematic reviews themselves will they become sufficiently critical of meta-analyses and the so-called “systematic reviews” carried out by others. Furthermore, such activities could benefit the systematic reviewers as well as their departments.48

Liver cirrhosis mortality does not seem to have changed significantly during the last 30 years in the Oxford region! How come? There are risks of erroneous findings (patient mix consequences, wrong sampling unit, diagnostic misclassification, type II errors), lack of beneficial effects of interventions considered to be effective, and incorrect implementation of interventions



  • Conflict of interest: None declared.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles