STOPAH

EXHIBITION

STOPAH: Prednisolone or Pentoxifylline for Alcoholic Hepatitis

Thursz. NEJM 2015; 372:1619-28. doi:10.1056/NEJMoa1412278

Clinical Question

  • In adult patients with alcoholic hepatitis, does prednisolone and / or pentoxifylline compared to placebo reduce mortality?

Design

  • Multi-centre, randomised, controlled trial
  • 2-by-2 factorial design with 4 treatment groups
  • Web-based electronic randomisation and assignment, with block sizes of 4 and stratification by geographical area and risk category (high vs intermediate)
  • Double-blinded (patients and clinicians / investigators)
  • Intention-to-treat analysis method
  • Power calculation:
    • 28-day mortality without drug 30%
    • 28-day mortality with drug 21%
    • Powered at 90%
    • Statistical significance level of 0.05
    • Patients required 1,026
    • Target recruitment 1,200 allowing 10% loss

Setting

  • 65 hospitals from across the UK
  • January 2011 to February 2014

Population

  • Inclusion: clinical diagnosis of alcoholic hepatitis
    • History of recent excess alcohol consumption
      • Men > 80g/day (10 units/day)
      • Women > 60 g/day (7.5 units/day)
    • Bilirubin > 80 µmol/l (4.7 mg/dl)
    • Maddrey Discriminant Function > 32
    • Age over 18 years
  • Exclusion:
    • Patients with severe kidney injury, uncontrolled sepsis, active gastrointestinal bleeding or requiring inopressor support were excluded unless stabilised within 7 days after admission
    • Jaundice > 3 months
    • Cessation of alcohol consumption > 2 months
    • Other cause of liver disease
    • Serum aspartate aminotransferase (AST) > 500 IU/l or serum alanine transaminase (ALT) > 300 IU/l
  • 5,234 screened –> 1,103 randomised
  • Baseline characters were similar between groups (data are mean ±SD of all patients):
    • Age: 48.7±10.2 years
    • Alcohol consumption:
      • women 149.5±104.3 g/day (18.7±13.0 units/day)
      • men 200.1±125.2 g/day (25.0±15.6 units/day)
    • Discriminant Function: 62.6±27.2

Intervention

  • Pentoxifylline 400mg + placebo
  • Prednisolone 40mg + placebo
  • Pentoxifylline 400mg + Prednisolone 40mg
    • All drugs prescribed for 28 days

Control

  • Placebo
    • Identical and prescribed for 28 days

Outcome

  • Primary outcome: mortality at 28 days was not statistically different between any individual group – p-value for drug interaction was 0.41
    • Prednisolone + placebo: 14.3%
    • Pentoxifylline + placebo: 19.4%
    • Prednisolone + pentoxifylline: 13.5%
    • Placebo: 16.7%

  • Logistic regression: adjusting for risk category and factorial design, demonstrated no statistically significant difference with either prednisolone or pentoxifylline individually
    • Prednisolone
      • Odds ratio: 0.72 (95% CI 0.52 – 1.01)
      • p-value: 0.06
    • Pentoxifylline
      • Odds ratio: 1.07 (95% CI 0.77 – 1.49)
      • P value: 0.69

 

  • Secondary outcomes:
    • 90-day mortality or transplant: no difference between groups
      • Prednisolone: OR 1.02 (95% CI 0.77–1.35); p-value 0.87
      • Pentoxifylline: OR 0.97 (95% CI 0.73–1.28); p-value 0.81
    • 1-year mortality or transplant: no difference between groups
      • Prednisolone: OR 1.01 (95% CI 0.76–1.35); p-value 0.94
      • Pentoxifylline: OR 0.99 (95% CI 0.74–1.33); p-value 0.97
    • Multivariate analysis of primary outcome: demonstrated benefit with prednisolone with statistically significant benefit
      • Prednisolone: OR 0.61 (95% CI 0.41–0.91); p-value 0.02; NNT 16
      • Pentoxifylline: OR 1.10 (95% CI 0.74–1.64); p-value 0.62
  • Adverse events:
    • Serious adverse events of infection: statistically significant difference demonstrating increased infections with prednisolone
      • Prednisolone 13% vs no-prednisolone 7%
      • Absolute risk increase: 6.01% (95% CI  2.47% to 9.54%)
      • Number-needed-to-harm: 17 (95% CI 10.5 to 40.5)
    • Serious adverse events leading to death: no statistically significant difference between individual groups or drug / no-drug comparisons

Authors’ Conclusions

  • Pentoxifylline, at 400mg for 28 days, did not improve mortality in this trial
  • Prednisolone, at 40mg for 28 days, may have a beneficial effect at 1 month but this was not demonstrated to persist to 90 days or 1 year

Strengths

  • Highly relevant research question
  • Pragmatic and efficient 2-by-2 factorial design
  • Good methodology with low risk of bias
  • Good follow-up in a population that can be difficult to research

Weaknesses

  • Diagnosis of alcoholic hepatitis was clinical rather than histological, which may be criticised by some — this is, however, a more pragmatic, ‘real-world’ approach
  • Power calculation over-estimated mortality — 16% actual compared with 35% expected — so effective power of this study was reduced and a false-negative conclusion is possible
  • Power calculation based on anticipated effect of a 30% reduction in mortality for each drug — is this a very optimistic aim?
  • Exclusion incidence and reasons not provided – did they need to exclude those with severe kidney injury, sepsis, GI bleeding or inopressor requirement? How frequently did this happen? This possibly limits generalisability.
  • ‘Statistical significance’ only demonstrated after extensive multivariate post-hoc analysis — this should not be considered a certain conclusion
  • Funding limitations meant that not all patients were followed-up for 90-day and 1-year outcomes — study terminated after last recruited patient followed-up to 28-day outcome

The Bottom Line

  • This study was under-powered to demonstrate clinically relevant effects
  • However, it suggests strongly with a low risk of bias that pentoxifylline 400mg daily for 28 days has no beneficial effect whilst prednisolone 40mg daily for 28 days may be beneficial for short-term outcomes but not for medium or long-term outcomes
  • From these data (with wide confidence intervals that cross the null hypothesis line)
    • Prednisolone NNT (28-day mortality) is 23
    • Prednisolone NNH (serious adverse infection) is 16

External Links

Metadata

Summary author: @DuncanChambler
Summary date: 5 May 2015
Peer-review editor: @DavidSlessor

Leave a Reply