Article Text

Download PDFPDF
127 Why does primary angioplasty not work in observational studies when it works consistently in randomised controlled trials? an analysis of 51242 patients with ST elevation myocardial infarction
  1. S Sen1,
  2. J E Davies1,
  3. I S Malik2,
  4. N Hadjiloizou1,
  5. R A Foale2,
  6. R Baruah1,
  7. G W Ghada2,
  8. A D Hughes1,
  9. J Mayet1,
  10. D Francis1
  1. 1Imperial College London, International Centre for Circulatory Health, NHLI, London, UK
  2. 2St Mary's Hospital, Imperial College Healthcare NHS Trust, London, UK


Introduction Meta-analyses of randomised controlled trials have shown long term mortality benefit of primary angioplasty (PPCI) over fibrinolysis for the treatment of ST segment elevation myocardial infarction (STEMI), making PPCI the standard of care for STEMI. Yet paradoxically, a recent meta analysis of observational studies appears to show reduced or no long term mortality benefit1; leading to the controversial possibility that primary angioplasty may not be beneficial in the ‘real world’. One explanation for these findings may be that patients with a high risk of mortality are preferentially allocated to angioplasty. We assessed this by determining the distribution of cardiogenic shock patients in such studies.

Methods The 11 observational registries in the meta analysis by Huynh et al were re-analysed. ‘High risk’ was taken by a study's own report of cardiogenic shock. For studies not reporting shock, either highest Killip class or systolic BP<100 mmHg were used. A χ2 test (Fishers exact, with Yates correction for larger studies) was used to determine if high risk patients were being allocated disproportionately to one therapy. Weighted least squares regression was used to determine the relationship between high risk patient distribution and apparent difference in outcomes.

Results In 10 of the 11 registries, there were data on cardiogenic shock or its equivalent (51 242 patients). In 7 of the 10 studies reporting sufficient baseline data, an excess proportion of shock patients were allocated to PPCI (8.4% PPCI vs 3.1% fibrinolysis, p<0.0001) (Abstract 127 Figure 1). In the other 3 studies (RIKS HIA, ACOS and ALABAMA), there was an excess of shock subjects allocated to fibrinolysis (3.9% in PPCI vs 12.2% in fibrinolysis, p<0.0001). Differences in outcome between PPCI and fibrinolysis was strongly linked to the excess allocation of cardiogenic shock patients with the arm with the higher proportion of cardiogenic shock patients tending to show with worse long term mortality (r=−0.92, p<0.01). Weighted regression indicated that for each extra percentage of shock patient allocated to an arm, outcomes in that arm were 6.9% worse than the other arm.

Abstract 127 Figure 1

The real world mortality difference between angioplasty and fibrinolysis depends on which arm high risk patients are preferentially allocated to. The arm allocated an excess of high risk patients tended to have worse outcomes, which explains the wide range of results. As is shown (arrow) without allocation bias angioplasty would be better.

Conclusion Data from “real world” observational registries of PPCI vs fibrinolysis should be interpreted with caution. Differential outcomes between arms are overwhelmingly driven by which arm the high risk patients are preferentially allocated to. This allocation bias can be dealt with statistically (arrow on Abstract 127 Figure 1) or, better, eliminated through randomisation. When this is done, PPCI remains the treatment of choice for STEMI in the ‘real world’.

  • primary angioplasty
  • myocardial infarction
  • fibrinolysis

Statistics from

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.