Effects of extended-release metoprolol succinate in patients undergoing non-cardiac surgery (POISE trial): a randomised controlled trial

Devereaux . Lancet 2008; 371: 1839–47 doi:10.1016/S0140- 6736(08)60601-7

Clinical Question

  • In patients undergoing non-cardiac surgery, do beta-blockers compared to placebo reduce the composite of death, MI or non-fatal cardiac arrest?


  • Randomised Controlled Trial of Metoprolol vs. Placebo
  • Computerised block randomisation stratified by centre, 1:1 allocation
  • Allocation concealed by using telephone service
  • Double-blinded
  • Intention-to-treat analysis
  • A priori power calculation with an assumed control event rate of 6%
    • A sample size of 10000 was calculated to give 92% power to detect a relative risk reduction of 25% (two-sided)
    • However acknowledged that 8000 patients would result in a still acceptable 85% power
    • In July 2007, without knowledge of the results, knowing that >8000 patients have been randomised and had a higher than predicted event rate, the study was terminated early as the study drug was about to expire
  • Robust monitoring including central data consistency, statistical monitoring and on-site monitoring (in centres recruiting ≥40 participants, or outliers on statistical monitoring)
  • Cox proportional hazards model used except for new clinically significant AF, CHF, hypotension, bradycardia or cardiac revascularisation, where Chi-squared test was used


  • October 2002 and July 2007
  • 190 hospitals in 23 countries
  • 8351 patients randomised (after exclusion of fraudulent data)


  • Inclusion: Patients with, or at risk of, atherosclerotic disease who were undergoing non-cardiac surgery
    • Aged 45 or older
    • Undergoing non-cardiac surgery
    • Expected hospital stay ≥24h
    • Plus one of the following:
      • History of CAD / PVD / Stroke
      • Admission for congestive heart failure in past 3 years
      • Undergoing major vascular surgery
      • Or had any 3 of the following 7 risk factors:
        • Undergoing intrathoracic or intraperitoneal surgery
        • History of CHF / TIA / Diabetes
        • Serum creatinine >175μmol/L
        • Age >70
        • having emergency or urgent surgery
  • Exclusion:
    • Heart rate <50bpm
    • 2nd or 3rd degree heart block
    • Asthma
    • Receiving a β-blocker already
    • CABG in the past 5 years and no ischaemia since
    • Low risk surgical procedure (clinician judgement)
    • Taking Verapamil
    • Prior adverse reaction to a β-blocker
    • Previous enrolment in POISE
  • 9298 patients randomised initially, 8351 left after exclusion of fraudulent data, 8331 completed 30-day follow-up
    • 8 patients in the Metoprolol group and 12 in the Control group were lost to follow-up
  • Very similar baseline demographics, co-morbidities, medications, surgery and anaesthetic techniques
    • Most patients (≈41% in each group) underwent Vascular surgery


  • Oral extended release Metoprolol or IV Metoprolol QDS if unable to take PO
    • Had to have HR ≥50 and Systolic BP ≥100mmHg to receive study drug
    • Pre-op dose (100mg Metoprolol PO) given 2-4h pre-op
    • Post-op dose (100mg oral extended-release Metoprolol) given as soon as HR ≥80 and Systolic BP ≥100mmHg, or at 6h postop
    • 12 hours after the first post-op dose, patients started taking 200mg oral extended-release Metoprolol once a day for 30 days
    • If HR consistently <45bpm or Systolic BP <100mmHg, study drug withheld until HR or BP recovered and restarted at half-dose (100mg)
    • IV administration for those unable to take PO, either slow (15mg over 60min) or fast (three doses of 5mg over 2 min) infusion 6 hourly


  • Placebo, administered using the same regimen and criteria as intervention


  • During monitoring, 6 hospitals in Iran and 3 hospitals in Colombia were found to engage in fraudulent activity and their patients were excluded from analysis. (Decision taken before trial results known)
  • Primary outcome: Composite outcome of cardiovascular death, non-fatal MI and non-fatal cardiac arrest at 30 days after randomisation
    • Metoprolol Group: 244 (5.8%)
    • Control Group: 290 (6.9%)
    • Hazard ratio (as reported): 0.84 (95% CI 0.70 to 0.99; P = 0.0399)
    • Absolute Risk Reduction (calculated): 1.10% (95% CI 0.05% to 2.15%; P = 0.0441)
    • Fragility index: 2
    • Number-needed-to-treat: 92
  • Secondary outcome: (data presented in table)
    • Death
      • Total mortality was greater in the metoprolol group
      • Non-cardiovascular deaths were greater in metoprolol group but did not reach statistical significance
      • Cardiovascular deaths were not different between groups
      • Cause of death was categorised as sepsis or infection more frequently in the metoprolol group
    • Cardiac events
      • Myocardial infarction (MI) was significantly less common in metoprolol group, especially the subgroup of non-fatal MI
      • Cardiac revascularisation (PCI) was significantly less common in metoprolol group
      • Congestive cardiac failure was not statistically significantly different but was more common in metoprolol group
      • New clinically significant atrial fibrillation was less common in metoprolol group
      • Clinically significant hypotension or bradycardia were both much more common in the metoprolol group
    • Cerebral events
      • Strokes were twice as common in the metoprolol group
      • Of those that suffered non-fatal strokes, the majority required assistance with everyday activities or were incapacitated, regardless of which group they were assigned
  • Subgroup analyses:
    • Pre-specified analysis of the primary outcome found the treatment effect was consistent across subgroups based on sex, type of surgery and use of epidural/spinal anaesthetic
    • Post-hoc analysis for mortality, MI and stroke with the above subgroups, plus new subgroups based on region, presence of on-site monitoring, presence of atherosclerotic disease identified no subgroup effect
  • Post-hoc multivariable analyses:
    • Clinically significant hypotension had the most significant link with death and stroke

Authors’ Conclusions

  • Perioperative use of extended-release Metoprolol reduces the risk of myocardial infarction, new clinically significant AF and cardiac revascularisation at the expense of significantly increased risk of death, stroke and clinically significant bradycardia and hypotension


  • Large, multi-centre, international RCT
  • Study design published in advance
  • Appropriate power even with early termination
  • Robust and effective monitoring, allowing the identification of fraudulent data
  • On-site monitoring in hospitals that collectively contributed 85% of all participants
  • Adequate concealment of allocation sequence
  • Blinding of patients, clinical staff and outcome adjudicators
  • Multiple meta-analyses by study team to place results in context
  • Multivariable analysis to attempt identification of mechanism of harm


  • Number of patients lost to follow-up is greater than the Fragility Index for the primary outcome
  • No data on effect of slow vs. fast IV administration vs. PO on heart rate and blood pressure
    • Fast IV administration may have very different haemodynamic effects to an extended-release oral preparation
    • This is important as hypotension was a potential mechanism of action for increased mortality and stroke
  • Relatively high rate of study drug discontinuation (temporary) at 18% vs. 12% with placebo

The Bottom Line

  • This large, high quality RCT is in agreement with previous studies and its results are supported by the meta-analyses
  • While different β-blockers or dosing regimens may have less adverse effects, based on this study, the routine introduction of β-blockers for the perioperative period in at-risk patients outside clinical trials, is not currently warranted
  • This study further underlines the importance of blood pressure control in the perioperative period, especially in patients receiving β-blockers

External Links

Editorial Comment

The publication of POISE 1 pre-dates the concerns raised regarding Don Poldermans’ research in this field. The meta-analysis performed in POISE 1 includes his work (DECREASE studies), about which there exists concerns of fraud and data accuracy. The systematic review linked to above has been published after Poldermans was fired from his academic post, and provides analysis with and without his research work. It is important to view the POISE 1 results within this context.


Summary author: Peter Szedlak
Summary date: 22 May 2017
Peer-review editor: Duncan Chambler

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.