Why we need a national registry in interventional cardiology

Percutaneous coronary intervention (PCI), comprising coronary angiography and angioplasty, has burgeoned as a less invasive means of revascularising coronary arteries than coronary artery bypass grafting (CABG). Stent insertion to maintain patency of the dilated artery and the substitution of drug-eluting stents (DES) for bare-metal stents (BMS) have been used to lower rates of restenosis. PCI has lessened the demand for CABG, and more patients with acute coronary syndromes (ACS) or limiting angina, who would not previously have been considered candidates for coronary revascularisation, are now undergoing PCI. In Australia, about 30 000 PCIs with stenting are performed each year, and the numbers continue to grow.1

However, it remains uncertain as to whether current PCI use reflects evidence-based medicine in maximising population health and return on investment. Resolving this issue in Australia may require the establishment of a national registry of PCI procedures.

No three-way randomised comparisons of the cost-effectiveness of PCI, CABG and medical therapy have been reported to date. Cost-effectiveness studies comparing routine management with conservative invasive management for NSTEACS suggest that the former is cost-effective, at about US$12 739 per life-year gained,21 but estimates are limited to the first 6 months after index hospitalisation and do not account for medium- to long-term events. In the case of stable coronary disease, CABG is definitely more cost-effective than medical therapy, which in turn is more cost-effective than PCI, even in “real-world” studies that have compared appropriate, guideline-concordant use of all three methods in unselected patients.22 Choosing DES rather than BMS renders PCI even less cost-effective, with estimated Australian costs up to A$46 829 per quality-adjusted life year (QALY), close to the accepted cost-effectiveness threshold of around A$50 000.23 Finally, cost-effectiveness estimates from different trials range widely, from US$25 540 per QALY gained (in the SIRIUS [Sirolimus-Eluting Balloon Expandable Stent in the Treatment of Patients with De Novo Native Coronary Artery Lesions] trial)24 to as high as US$128 744 (in the BASKET trial).25 Cost-effectiveness in most trials has probably been overestimated owing to underestimation of stent use rate per patient (as multi-lesion stenting becomes more prevalent), overestimation of the reduction of TVR by DES, and failure to measure downstream costs of late complications due to LST and bleeding induced by dual antiplatelet therapy.18

Maximisation of clinical benefits and cost-effectiveness will be more likely if PCI is performed preferentially on patients with trial-validated indications and high baseline risk of death or future coronary events: those with early-phase STEMI, intermediate- to high-risk NSTEACS, or severe limiting angina due to multivessel disease refractory to optimal medical therapy and otherwise unsuitable for CABG.

However, US registry data indicate that about 85% of all PCI procedures are undertaken electively in patients with stable coronary disease.26 In the minority of patients with STEMI undergoing primary PCI, this occurs within 90 minutes in fewer than 5% of patients transferred from non-PCI-capable facilities,27 and in general PCI is used more frequently in low-risk than high-risk patients.28 Surveys of invasive cardiologists in the US suggest a bias towards performing PCI in situations lacking evidence of benefit,29 such as treating stable ischaemia,8 opening infarct-related occluded arteries,6 and intervening on multiple as opposed to single “culprit” lesions.30 Any potential overuse of PCI is being matched with an underuse of effective medical therapy: no more than 50% of eligible patients with NSTEACS in the GRACE (Global Registry of Acute Coronary Events) registry received all indicated therapies.31

DES accounted for 85% of all stent insertions to 2006 in the US, at least half of which were performed for what were, at the time, “off-label” or untested indications in higher-risk patients, including those with diabetes.32 Such patients had coronary lesions more prone to restenosis than those in pivotal trials: long lesions, chronic or small-vessel occlusions, calcified or bifurcation lesions, and vein-graft lesions. The benefits, if any, of using DES over BMS for off-label indications are only now being assessed in trials and registries which, to date, feature small numbers of patients, surrogate outcomes (such as late luminal loss and restenosis), results yet to be published in peer-reviewed journals, or no definitive differences in rates of major adverse cardiac events (MACE).33 Data from three registries have shown higher rates of TVR at 12 months in off-label indications than in standard indications in patients receiving DES (7.6% v 4.4%,34 6.3% v 2.4%,35 and 12.7% v 7.7%;36 P < 0.001 for all comparisons), although the most recent report indicates significantly lower mortality with DES than with BMS for off-label indications (3.7% v 6.4%; P < 0.001).36

The registry literature comparing DES with BMS is littered with conflicting reports. For example, a large Swedish registry study reported an absolute increase in mortality of 1.5% at 3 years in DES patients,37 but this was no longer seen at 4 years.38 A Swiss study over 3 years also reported 1.5% excess mortality but with no apparent flattening of the cumulative stent thrombosis curve over time, suggesting ongoing risk.39 Another observational study indicated that LST risk could be two- to threefold higher, up to 1.3% per year.14 In contrast, a Canadian registry study to 2 years, in which patients received clopidogrel for 12 months after PCI, suggested no DES-related excess in stent thrombosis, MI or mortality compared with BMS,40 and a US study of patients with both simple and complex lesions showed significant reductions in adjusted rates of TVR and MACE in favour of DES.41 As of mid 2008, the cumulative observational evidence is increasingly reassuring that prevention of restenosis achieved by DES does not come at the expense of an increase in death or MI compared with BMS.42,43

The disparity in results from trials and registry studies may relate to differences in study enrolment criteria (for both patients and coronary anatomy), definitions of stent thrombosis and other clinical events, clinical thresholds for repeating TVR, number of stents inserted, adequacy of stent placement, type and duration of post-PCI antiplatelet therapy, and follow-up period. The risk of PCI-related bleeding in routine practice also needs closer scrutiny with the advent of more potent antithrombotic agents, with registry data reporting a 5.5% incidence of major periprocedural bleeding associated with a 64% increased risk of inhospital death in patients undergoing PCI for ACS.44 More data are needed on the risk of bleeding associated with long-term use of dual antiplatelet therapy — particularly in patients receiving warfarin prophylaxis for atrial fibrillation (in whom major bleeding rates as high as 7% per year have been reported)45 — and the risk of re-infarction if such therapy needs to be discontinued at the time of surgery.

Given the uncertainties around efficacy, safety and cost-effectiveness, what is the current state of play of PCI use in Australia in 2008? Frankly, we have few data from which to get a clear idea. According to the most recent statistics, the annual number of PCI procedures has more than doubled from 80 per 100 000 population in 1996–97 to 170 per 100 000 in 2005–06,46 with significant variation in per-capita rates between states. DES use has increased exponentially, accounting for virtually all stents inserted in the private sector in 2006 and about 30% in the public sector.

However, the last detailed analysis of PCI in terms of indications, patency rates, survival and complications was published in May 2002, based on data collected within the Coronary Angioplasty Register between 1980 and 1999.47 In that period, 42% of angioplasties were performed in patients with stable angina, 42% in patients with unstable angina, and 7% as either primary or rescue angioplasty for STEMI. Virtually all procedures (92%) reported in 1999 involved BMS (as DES were not yet available) and more than 85% targeted single vessel disease, among which 38% of lesions affected the left anterior descending artery. The vast majority (95%) of procedures were judged successful; the remainder involved complications requiring urgent CABG in fewer than 1% of cases, peri-procedural acute MI in 1.2%, other arterial complications such as stroke in 1.2%, and death in 0.8%. While these data appear reassuring, 10 of 57 hospitals (accounting for 15% of procedures) did not provide data, and those that did provided incomplete data for some variables.

In a more contemporary national cohort of 3402 patients presenting with ACS to 39 centres (29 PCI-capable), angiography was undertaken in 57% of patients, of whom half underwent PCI, and 57% of those received DES.48 The timing of these procedures and long-term outcomes were not reported. This audit also showed that only 58% of patients with STEMI, 35% with non-STEMI and 24% with unstable angina received all five secondary prevention therapies at discharge (aspirin, clopidogrel, β-blocker, angiotensin-converting enzyme inhibitor or angiotensin-receptor blocker, and statin). Among 2428 cases of PCI performed with stent implantation in seven Melbourne hospitals between April 2004 and December 2005, DES was used in 45% of cases, of which 88% involved patients with diabetes, small vessels, long lesions, in-stent restenotic lesions, chronic total occlusions and bifurcation lesions.49 The indications and outcomes for PCI in this study are yet to be reported.

Although randomised trials will (and should) remain the gold standard for determining PCI efficacy, extrapolating trial results to routine practice in interventional cardiology is limited by:

• selection bias, with the frequent exclusion from trials of high-risk patients with complex disease, who may benefit from PCI;

• inferior comparators as rapid changes in PCI technology outpace fixed protocols of trials that may span several years;

• insufficient power of trials to discern variations in PCI effects among different patient subgroups or clinical settings, or to identify long-term hazards;

• limited generalisability of results achieved in PCI centres of excellence participating in trials; and

• clinician misperception of benefit due to conflation of patient-critical endpoints (avoidance of death or re-infarction) with quality-of-life endpoints (avoidance of TVR as a marker of angina-free status) within composite outcomes.

A strong case can therefore be made for a large-scale, prospective registry of PCI that is able to longitudinally track patient outcomes in relation to baseline patient characteristics, indications for intervention, coronary anatomy, procedural technique and site of care (Box). Undertaken at a national level using standardised definitions and datasets, such a registry could detect trends towards adverse outcomes and variations in PCI use that may warrant reassessment of its use in specific circumstances (as a form of “post-marketing surveillance”) and facilitate monitoring and benchmarking of PCI performance. Calls have been made for the establishment of a national cardiac procedures registry,50 but, apart from preliminary work in data definition, no further action has been taken.

There is an urgent need for both government and professional advocacy in this area: governments need to provide infrastructure support and funding for the establishment and maintenance of such registries; and interventional cardiologists need to design and populate the registry with complete and accurate datasets, and alter practice according to its findings. Expecting patients and funders to continue supporting unrestricted and unmonitored use of costly PCI in situations where efficacy and safety concerns remain unresolved is clearly undesirable, and the creation of a national PCI registry should now proceed without further delay.