The CARDia trial protocol.

An earlier edition of the web based version of Heart (May 2003) saw the electronic publication of the protocol of the CARDia (coronary artery revascularisation in diabetes) trial. The CARDia trial is a multi-centre, “prospective, randomised comparison of optimal coronary angioplasty—with use of stenting and abciximab recommended—versus up to date coronary artery bypass grafting in patients with diabetes mellitus suitable for either intervention”. It will recruit 600 patients through 20 centres over a two year period. Patients suitable for either procedure will be randomised to percutaneous coronary intervention (PCI) or coronary artery bypass grafting (CABG). The PCI group will be further randomised to receive bare stents or rapamycin eluting stents. The publication of a trial protocol is a new departure for Heart but follows precedents in other journals when important trials are planned or underway. It might at first sight seem odd to publish protocols of trials in peer review journals before the trial has completed and the results are available. After all a trial protocol without the results is little more than a well thought out idea and as such has little more value than reading the product of a bar-side brainstorming session among colleagues. Where will it end—will we see publication of lists of ideas that bubbled up at cardiologists’ bath time? Clearly the gravity, relevance, and importance of a trial will determine whether the protocol should be published on its own. Does the CARDia trial meet such criteria and warrant web space? I believe it does. At rates of 15–20%, patients with diabetes form a very significant proportion of those undergoing revascularisation procedures and they are at considerably higher risk than non-diabetics whether they undergo percutaneous or surgical revascularisation. The five year survival for PCI in diabetics of 88% versus 93% in non-diabetics is mirrored in the surgical experience reported by Szabo and colleagues of 84.4% versus 91.3%, where the increased postoperative morbidity is also highlighted. With such high prevalence and risk it behoves us to be absolutely clear about optimal management particularly with regard to revascularisation strategy. Yet current management strategies are based on generalisations supported by retrospective subgroup analyses of out of date trials. O’Neill made the much quoted statement that “Multivessel angioplasty in diabetics should be abandoned” after the publication of the diabetic subgroup analysis of the BARI trial revealed a significantly higher five year mortality in the PCI group. The trial, and possibly the statement, have had a significant effect on clinical practice. A brief review of the nine PCI versus CABG trials over the last 15 years shows how unlikely it is that such analyses have value. The proportion of diabetics recruited varies from 6% in the RITA trial to 27% in EAST and in all but the very latest of these trials, the SoS trial, the diabetic subgroup was not prespecified. Of the multi-vessel balloon angioplasty versus CABG trials, the EAST trial showed no significant difference in mortality at five years compared to the 34.5% PCI versus 19.4% CABG (p < 0.003) reported in the BARI trial. In the stent era the three trials ARTS, ERACI-2, and SoS show no significant difference in mortality between multi-vessel stenting and CABG in diabetics, but there the similarity between the trial results ends. In ARTS there is a significant increase in death in the diabetics undergoing PCI (6.3%) compared with the non-diabetics (1.6%, p < 0.011), although this is not reflected in the SoS results where there is no difference. Indeed in the SoS trial, the only trial with diabetes as a prespecified subgroup, the presence of diabetes has no effect on death or revascularisation rates at all (table 1). Are we to believe the SoS trial because the diabetics were a prespecified subgroup? Fourteen per cent (142) of patients recruited were diabetic so the numbers are still relatively small. We are all aware that subgroup analysis should be viewed with great caution, but the erroneous effects are probably underestimated even in prespecified subgroups.