By John Kellett, MD
Several large trials have shown that fibrinolytic therapy can improve survival after acute myocardial infarction (MI). However, fibrinolytic drugs also increase the risk of a major bleeding episode or stroke, leaving the physician with a potentially difficult choice - to give fibrinolytic therapy and take the risk of precipitating a stroke, or withhold it and take the risk of preventable death from MI. The decision to administer fibrinolytics would not be too difficult if all patients were similar to those who fulfilled the admission criteria for the large clinical trials: that is middle aged, with definite acute MI, presenting shortly after symptom onset and with no particular risk factors for stroke or bleeding, and if not fibrinolysed, would have a mortality of around 10%. Unfortunately, such patients may not always be encountered by physicians in everyday practice. In real life, both the time of symptom onset and the diagnosis of MI are often far from certain, and individual patients may have hypertension, low body weight, advanced age, or other factors that expose them to increased risk of bleeding, stroke, and death. It is not surprising, therefore, that the proportion of patients with acute infarction actually given fibrinolytic therapy may be as low as 20%, and that up to 6% of patients without infarction may be inappropriately fibrinolysed. It is possible however, to determine explicitly the risks and benefits of fibrinolysis by decision analysis, provided the probabilities of MI, death from MI, and stroke can be estimated.
As far as I know, Nenagh CCU is the first to use a decision analysis model to guide fibrinolytic therapy (Clin. Cardiol. 1998;21:93-98). We have used two approaches to determine the chance of acute MI: one based on the patient's presenting ECG, and the other on a logistic regression model. Both give excellent, and comparable results (Canad. J. Card. 1997;13:1033-38). Logistic regression models are also used to determine the chance of stroke (Circulation 1995;92:2811-18) and death from infarction (Med. Care 1992;29:1196-1211).
The program printout does not state explicitly whether fibrinolytics should be given or not. However, it provides all the data needed to make this decision. These include the chances of death with and without treatment, as well as the "up front" risks of stroke and serious bleed. It then estimates how long the patient is likely to live with and without fibrinolysis. It also performs multiple sensitivity analyses to determine how much greater or smaller the chances of infarction, stroke, death, bleeding, time of onset etc., would have to be before the preferred treatment option changes. The physician can decide if he thinks the likely life expectancy gain from fibrinolysis is worth the "up front" risk of stroke. The physician can also decide, based on the data provided, which fibrinolytic regimen should be used. Accelerated tPA is not only riskier than streptokinase, it is also substantially more expensive. In Nenagh CCU, we have decided not to give tPA is it costs more than IR£10,000 per extra year of life saved. This not because we do not think our patients' lives are worth less than IR£10,000 per year, but because once the cost-effectiveness of tPA over streptokinase reaches this level, it starts to rise very rapidly in a very short time. Thus, by the time the accelerated tPA regimen is set up it is no longer cost-effective.
Development of the program started in 1989,
and was first reported in 1992 (MD Computing 1992;3:157-164). The decision
tree (Med. Decis. Making 1995;15:297-310) has been constantly updated since
first published in 1991. During the early development of the program CCU
staff were blinded to its results (i.e. there was no printout). Of the
262 patients treated during this phase the overall infarct mortality was
21.4%, and 2.2% of patients without infarction were inappropriately fibrinolysed.
As of September 1998, 518 patients have been managed with decision analysis
support, with an overall infarct mortality of 13.9%, and only 0.9% of those
without infarction inappropriately fibrinolysed. Confirmation of these
encouraging findings will require a larger study performed over a shorter
time, thereby ensuring that the results are not confounded by other changes
in CCU management. I would welcome collaboration with any interested colleague(s).
Demonstration versions of the program are available from me via e-mail
at kellett@iol.ie.