Statistics from Altmetric.com
The enthusiasm of researchers and clinical investigators is an admirable and perhaps necessary trait for performing good clinical research but must not obscure the interpretation of trial results. Cost-effectiveness analysis is being promoted as an objective means to measure the balance between the effectiveness of a treatment and its cost. Although cost-effectiveness analysis in medicine is in its infancy, much work has been performed to solidify its methodological underpinnings.1-3 Clinicians, nevertheless often remain suspicious of the process as they realise how difficult it may be to have reliable cost and efficacy information.
This issue contains an article by Aristides and colleagues on the cost-effectiveness of abciximab in preventing restenosis following percutaneous transluminal coronary angioplasty (PTCA).4 We agree wholeheartedly with the caveats expressed by Glasziou in his accompanying editorial5 and add some further worries in this commentary.
The key ratio in cost-effectiveness analysis has a numerator of costs and a denominator that measures clinical efficacy such as years of life saved. In assessing the efficacy of an intervention it is necessary to consider critically the totality of the evidence. Aristides et al have relied only on the data from the 2009 patients enrolled in EPIC,6 7 although referring to the results of the more recent EPILOG8 and CAPTURE9 trials to support their argument for the safety of abciximab. However, these two trials address a reasonably similar patient population to EPIC and therefore the 4057 patients included in these later trials could and should have been incorporated into the measure of efficacy.
In EPIC there was no mortality advantage nor any additional protection from myocardial infarction beyond what was observed at 30 days. The 8.1% absolute decrease in the composite end point at six months was largely due to a decrease in repeat PTCA (20.9% v14.4%). However, only US centres participated and the American proclivity for aggressive interventional cardiology is well appreciated.10-12 For example, a recent population study of elderly postmyocardial infarction patients showed that PTCA is performed eight times more frequently in American than Canadian patients with no difference in mortality at one year.12 It is therefore quite possible that a soft end point, such as repeat PTCA, may be practice driven. The possibility of important geographic practice variations affecting outcomes, coupled with international variations of costs, invalidates the conclusion of Aristides et al that their economic results are transferable to other health care systems.
How can one reconcile the diverse clinical results? EPIC recruited patients from 1991 and recent improvements in patient selection and equipment, especially the introduction of stents, may have influenced the impact of abciximab. In addition, the maximum efficacy of abciximab in EPIC was observed in patients undergoing direct coronary atherectomy,13 a technique not widely used outside a few specialist centres.
The totality of this evidence thus suggests that abciximab following PTCA has no significant effect on mortality nor on repeat revascularisations at six months, but may decrease early myocardial infarction, particularly non-Q wave infarction, by about 40%. These conclusions are supported by the prevailing evidence of the importance of platelets in thrombosis whereby early post-PTCA complications are due to platelet activation and subsequent aggregation. Restenosis is a more intermediate term complication resulting from vessel recoil and smooth muscle proliferation, while late complications are often due to disease progression in other vessels.
Aristides et al present economic analyses that are less than perfect; the estimate of clinical efficacy suffers from selection bias and the Markov model is also limited by the selective use of historical controls from the 1980s that is particularly inapplicable given the rapid advances in PTCA technology. To gain the trust of both physicians and administrators, the parameters used in this model need to be stated more explicitly. The decision tree for the model should be presented as well as the probabilities assigned to each branch.
The economic analysis relying exclusively on the EPIC data is more transparent but appears to have an accounting error. The incremental cost of $1054 is reported to be associated with an incremental efficacy at six months of 8.1% to give a cost of $13 012 (1054/0.081) per patient free from a serious event. However, during the initial hospitalisation there were 4.5% fewer events with abciximab and the costs of these events had already been considered in calculating the incremental cost. Consequently, the incremental cost of $1054 is actually associated with an incremental efficacy of 3.6% (8.1% − 4.5%) giving a cost of $29 000 (1054/0.036) per patient free from a serious event.
Owing to these shortcomings, we present our own “back of the envelope” economic analysis. Current myocardial infarction rates following PTCA are approximately 3–5% and if the efficacy of abciximab to reduce myocardial infarction is 40% then about two infarctions will be avoided for each 100 patients treated, at a total cost in the vicinity of $150 000. As a first approximation, the impact of myocardial infarction may reduce a patient’s quality of life to 0.9 (1 being full health). If one assumes a 10 year survival following PTCA, one could expect that the prophylactic treatment of all high risk patients would add two quality adjusted life years (QALYs) at a cost of $75 000/QALY. By conventional technology assessment standards this would place abciximab in a relatively unfavourable position.14 However, the adoption of a more selective programme treating only patients with intraluminal thrombus identified at arteriography may be more cost effective. This approach is analogous to the increased cost effectiveness of secondary compared to primary prevention for hypercholesterolaemia.
Nevertheless the glycoprotein (GP) IIb/IIIa agents are an exciting class of drugs and the accumulation of further data may establish their importance in acute coronary syndromes. This role will likely vary in different patients in different health care systems. As more compounds are introduced it may be anticipated that costs will decrease. However, the true costs of the benefits of these drugs must be fully appreciated so that informed societal decisions may be taken.
This commentary is meant to guard against “GP IIb/IIIa mania”. Regarding economic analyses, we propose the systematic inclusion of all relevant studies to estimate best the parameters of interest. The assumptions and data employed in any mathematical model must be transparent and understandable to the clinicians and administrators who will use the analyses to guide their decisions. Extreme care must be exercised if one attempts to transfer cost-effectiveness analysis across international borders. As with the general medical literature, readers must critically examine cost-effectiveness analysis for their applicability in their own environments.