Article Text

Download PDFPDF

Drug eluting stents: maximising benefit and minimising cost
  1. J Gunn1,
  2. A C Morton1,
  3. C Wales1,
  4. C M H Newman1,
  5. D C Crossman1,
  6. D C Cumberland2
  1. 1Cardiovascular Research Group, Division of Clinical Sciences (Northern General Hospital), University of Sheffield, Sheffield, UK
  2. 2Ampang Puteri Specialist Hospital, Jalang Memanda, Mukin Ampang, Selangor, Malaysia
  1. Dr Julian Gunn, Cardiovascular Research Group, Division of Clinical Sciences (Northern General Hospital), University of Sheffield, Clinical Sciences Building, Northern General Hospital, Sheffield, S5 7AU, UK;


A policy of selective implantation of drug eluting stents, in a minority of lesions most likely to benefit, seems to be a rational way to employ this new and currently costly technology

  • drug eluting stents
  • sirolimus
  • percutaneous coronary intervention
  • restenosis
  • CABG, coronary artery bypass grafting
  • IVUS, intravascular ultrasound
  • MLD, minimum lumen diameter
  • PCI, percutaneous coronary intervention
  • RAVEL, randomized comparison of a sirolimus-eluting stent with a standard stent for coronary revascularization
  • SIRIUS, US multicenter, randomized, double-blind study of the sirolimus-eluting stent in de novo coronary lesions
  • TLR, target lesion revascularisation
  • TVF, target vessel failure

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.

The RAVEL and the SIRIUS studies, using a sirolimus eluting stent, report substantial reductions in clinical restenosis with drug eluting compared with bare stents. The drug eluting stent has been hailed as the third great breakthrough in percutaneous coronary intervention (PCI) after the balloon and the stent, but excitement greeting its arrival must be tempered by three facts. Firstly, it is impossible to predict exactly which individual coronary artery stenosis will restenose (making it difficult to know who would benefit from the new stent). Secondly, “real world” PCI includes many lesion types more adverse (and probably more deserving of antirestenosis strategies) than those included in the studies. Thirdly, the first drug eluting stent on the market costs about five times more than a conventional stent (making it economically impossible to treat all lesions). We need to decide how to make rational treatment decisions with this new device within limited health budgets. In this article, we attempt to do exactly that.

The first drug eluting stent (Cypher, Cordis) has recently been made commercially available in the UK. Data from the RAVEL study and the “first-in-man” studies suggest that this device, which elutes the drug sirolimus (rapamycin, a macrolide antibiotic, with immunosuppressant, antiproliferative, and antimigratory properties), produces a zero restenosis rate up to, and beyond, one year in selected lesions.1–3 Interestingly, the larger SIRIUS study (similar to RAVEL, but performed in the USA), as yet unpublished, has complete follow up data available for the first 400 patients enrolled; the study demonstrates (as its primary end point) a 12% target vessel failure (TVF) rate in the sirolimus group compared with 23% in the control group, while the target lesion revascularisation (TLR) rate (an index of which restenotic lesions really matter) was 6.8% versus 21.5%—a two thirds reduction.4 The growth of in-stent neointima in this study was, as in RAVEL, negligible. The relatively high TVF rate is being explained as angiographic restenosis beyond the ends of the stent, perhaps caused by balloon deployment there. If this simple procedural (rather than biological) explanation is true, we can expect a substantial reduction in (but not necessarily total abolition of) in-stent restenosis with drug eluting stents.

The impact of a device which promises to reduce restenosis (relative to a conventional stent) more than the Palmaz-Schatz stent did (relative to a balloon) can, therefore, be easily imagined. It is a situation analogous to that which developed in the mid 1990s, when use of the Palmaz-Schatz stent reduced angiographic restenosis from 32% after balloon angioplasty to 22% in the stent group, and the re-PCI rate from 21% to 10% (in highly selected lesions).5 As a result, interventional cardiology moved into the “stent era”, stents being implanted in 70–90% of PCI procedures (in decidedly unselected lesions). It is natural that interventionists will want coated stents, and patients will demand them. This enthusiasm, however, comes with a cost. The Cypher drug-eluting stent is being marketed in the UK at a price about fivefold that of a bare stent. This mandates a vigorous appraisal of the financial implications of its use, at a time of expansion in PCI, in an increasingly cost conscious health service.


It is often stated that it is not possible to predict which patients will undergo restenosis. It is, however, true that certain lesion (and, to a lesser extent, patient) characteristics are associated with higher restenosis rates. Multivessel stenting and diabetes are the two main patient related variables increasing the probability of restenosis. In the case of diabetes, the relative risk is about 1.3:1.2 The evidence for specific lesion locations having high restenosis rates is generally unconvincing (such as the proximal left anterior descending artery6,7), an exception being the saphenous vein graft.8 Angiographic, lesion related variables are more important, notably the minimum lumen diameter (MLD) pre-procedure,6,7 the MLD post-procedure.6,7,9 the acute gain,8 the vessel size (reference diameter),6,7,9,10 stent length,7,9,11–13 and the presence of multiple stents (probably a manifestation of the same phenomenon).9,12 Intravascular ultrasound (IVUS) based studies support these observations and show that an ostial location, plaque burden (plaque area/arterial area) and final lumen dimensions (MLD and lumen cross sectional area) are also important.14 Approximately half of patients with angiographic restenosis (diameter stenosis > 50%) present with clinical relapse.5 These usually require repeat revascularisation, the majority by PCI. On the other hand, “real world” restenosis rates are likely to be higher than those found in randomised controlled trials. Up-to-date, local data are important. Examination by one author (JG) of the records of 219 PCIs performed in 2001 (an unselected series, including acute and chronic cases, and all types of lesion) revealed that a mean of 1.6 vessels per patient were treated with a mean of 1.1 stents per vessel. The clinical restenosis (TLR) rate for these patients was 10.0% (of which 8.6% were for restenosis and 1.4% re-restenosis), 96% being treated with re-PCI and 4% with coronary artery bypass grafting (CABG). This figure of 10% TLR is very similar to those seen in unselected, contemporary, published series from other centres and operators.15


On the one hand, we could simply implant a drug eluting stent in every patient. This would substantially reduce the incidence of restenosis (perhaps by about two thirds, if we believe SIRIUS), and require no effort at predicting the most likely lesions to restenose. Initially, however, this approach would result in an approximately fourfold increase in stent budget (even with a discount in purchasing the drug eluting stents). A more responsible approach might be to identify a subset of lesions at the highest risk of restenosis, and target these for implantation of a drug eluting stent. This might yield a useful (though less substantial) reduction in restenosis at considerably less cost. In this paper, we aim to create a practical system for establishing different levels of risk of restenosis, and examine the impact of implanting drug eluting stents in lesions at different levels of risk.


Of all the risk factors for restenosis discussed above, we considered to be most useful “real world”, lesion based risk factors that were readily prospectively measurable in the majority of patients undergoing PCI. We therefore rejected IVUS based variables and confined ourselves to simple, practical, quickly assessed, angiographic, lesion based parameters. We focused on intended stent width (≈vessel reference diameter) and length (≈7 mm longer than the lesion). MLD pre-procedure was rejected as being too difficult to assess “by eye”, and MLD post-procedure was rejected because it is post hoc. We obtained lesion based restenosis data from the literature.6,7,9,10,12,13 Extrapolation and estimation were necessary (because some papers considered the impact of stent length, while others considered the impact of reference diameter, and each used different stents). The only patient based variable important enough to be included in the analysis was diabetes. The risk of restenosis associated with multi-lesion stenting was assumed to be the sum of the risk of each lesion. We also assumed reasonable (but not exceptional) standards of stent implantation (a residual stenosis in the range 10–15% is found even in well conducted studies3). We also determined the number of stents of each width and length used in contemporary practice from our catheter laboratory records. This was important to establish which stent sizes contributed most to the restenosis “burden” to the hospital in terms of sheer numbers.


We plotted the angiographic restenosis rate, predicted by the literature, for different stent diameters (≈reference diameter, divided into 0.25 and 0.5 mm groups) and lengths (≈lesion + 7 mm, divided into 5 mm groups)6,7,9 in table 1. We made the assumptions that the percentage increase in restenosis risk with increasing stent length applies proportionately to all stent widths, and that the percentage increase with decreasing stent width applies proportionately to all stent lengths. The per cent of each size of stent used in our hospital, broken down by the same criteria of length and width, was also plotted on this table. Table 1 reveals that stents of smaller calibre and longer length are at greater risk of restenosis—for example, a 2.5 × 20 mm stent would be expected to have an angiographic restenosis rate of about 32%. Significantly, it also shows how rare restenosis is in stents which are short and large in calibre—for example, a 4 × 8 mm stent would be expected to have an angiographic restenosis rate of about 2%. Table 1 also shows that there are relatively few stent sizes with a high risk of restenosis which are frequently used. When using table 1 it is important to remember that the figures quoted are merely a guide to the expected chance of angiographic restenosis. For the average patient (> 1 lesion with > 1 stent) the composite angiographic restenosis rate is the sum of that of all the stents deployed and the clinical restenosis rate is about half of this composite figure.

Table 1

Angiographic restenosis rates for different widths and lengths of bare stents


Application of three arbitrary thresholds of lesion based restenosis risk (> 5%, > 10%, and > 15%) to table 1 identified stent sizes with the highest priority for implantation of a drug eluting stent. For a 15% restenosis threshold, few groups of stent sizes were found that were actually used (the others either had a low restenosis rate (large stent widths) or were hardly ever used (long stents with small widths)). The potentially important sizes, for this threshold, were 2.5 × 10–20 mm, 2.75 × 15–25 mm, and 3.0 × 25–30 mm (plus 3.0 × 15–25 mm for diabetics, who comprise about 20% of the PCI population). From the manufacturer’s list of available sizes, we would require 2.5 × 13, 2.5 × 18, 2.75 × 18, 2.75 × 23, 3.0 × 18, 3.0 × 23, and 3.0 × 28 mm sized Cypher stents. The numbers of these required per annum, expressed as percentage of the entire stock of stents, would be 18%, distributed in the ratio 21:8:12:1:40:3:15, respectively (table 2). For the 10% threshold, using tables 1 and 2 in a similar way, the proportion of drug eluting stents would be about 38% and, for the 5% threshold, about 77%.

Table 2

Sizes of Cypher drug eluting stent marketed by Cordis


Multiplication of the per cent restenosis risk for each stent size category by the per cent of stents used in each size category (that is, the two figures in each box in table 1) yields the “restenosis burden” of each stent size. For example, the greatest restenosis burden (130/903 = 14%) is provided by stents in the category 3.0 × 15–20, because they are numerous and have a moderate restenosis rate; whereas 2.5 × 25–30 mm stents have a low restenosis burden (4%) because, despite a high restenosis rate, they are rarely used. Addition of the figures for restenosis burden, for all stent sizes, reveals an overall, stent based, restenosis rate of 9.0%. This stent based, angiographic restenosis rate, derived from independent data, is consistent with our own, “real world”, clinical restenosis rate of 10% (for 1.6 vessels per patient and 1.1 stents per vessel). Assuming that the stent sizes selected for drug eluting stent implantation have a restenosis rate about one third that of conventional stents (see first paragraph above), we can examine the impact upon restenosis of implanting drug eluting stents at selected thresholds of expected restenosis (first part of table 3). Application of a 15% restenosis risk threshold would result in reduction of this overall restenosis rate to 6.9%—a 23% relative reduction. For the 10% threshold, the overall restenosis rate would fall to 5.4%—a 40% relative reduction. For the 5% threshold, the overall restenosis rate would fall to 3.7%—a 59% relative reduction. It would be reasonable to expect proportionately similar reductions in the patient based restenosis rate (from 10% to 7.7%, 6.0% or 4.1%, respectively). Table 3 clearly shows that policies invoking progressively higher rates of implantation of a drug eluting stent exhibit a law of “diminishing returns” in terms of reduction in restenosis.

Table 3

Sample figures for selection of drug eluting stents at different thresholds of restenosis risk


Stent prices vary from centre to centre and from country to country. Making some reasonable assumptions, and knowing the price of the Cypher quoted to us by Cordis, together with their discount levels for increasing numbers of stents ordered, however, allows us to construct some realistic estimates of the economic impact of various thresholds of drug eluting stent implantation (second part of table 3). Application of a restenosis risk threshold of 15%, requiring 18% drug eluting stents, at the sizes listed above (with a 19% price discount) would require an increase in the stent budget of about 55%. A restenosis threshold of 10% (with a 22.5% discount) would cost an extra 110%; and a threshold of 5% (with a 22.5% discount) an extra 224%. Policies invoking progressively higher rates of implantation of a drug eluting stent would, therefore, demand steep increases in total stent costs. On the other hand, a reduction in restenosis (and re-restenosis) would reduce the number of repeat PCIs. Notionally, this would result in attenuation of the extra cost associated with using the new stents (also shown in table 3). However, in the “real world”, where expenditure is measured by the number of cases performed per year (not the number of patients), this would be less apparent, because patients with restenotic lesions would be replaced by those with de novo lesions. In these calculations, the number of other complications (CABG, myocardial infarction), which can add significantly to the costs created by an individual patient, are assumed to be few in number and equally distributed between conventional and drug eluting stents.4


Would the increase in stent expenditure resulting from the proposed use of a drug eluting stent be better spent on increasing the number of (bare stent) PCIs? For a centre performing 700 PCIs per annum, current practice, with a 10% clinical restenosis rate, allows 630 patients to be treated, the remainder of the procedures being repeat PCIs for restenosis. Use of the Cypher stent in lesions with > 15% risk of restenosis, at a cost of an extra 55% in stent expenditure, with a reduction in the overall clinical restenosis rate from 10% to 7.7%, would increase this number to about 646; a gain of 16 patients. The same money could, instead, purchase approximately an extra 72 PCIs with conventional stents which, with a “conventional” restenosis rate of 10%, amounts to 63 patients. This highlights nicely the dilemma; to balance moderately increased “quality” (clinical restenosis reduced from 10% to 7.7%) and slightly increased “quantity” (an extra 2.5% patients treated in a year) against the possibility of no change in quality but considerably better quantity (an extra 10% patients treated in a year). The greater the proposed expenditure on the drug eluting stents, the greater this difference; a few more cases of restenosis being saved, in contrast to the potential for many more de novo patients treated. In a financially constrained world, one could argue that the latter is preferable, provided that re-PCI is performed at low risk.


What will be the impact of using drug eluting stents in multi-vessel disease? The main reason for not performing multi-vessel PCI is no longer safety but restenosis. The main difference between multi-vessel stenting and CABG is the increased incidence of repeat revascularisation seen in the stent group.16 In simple terms, the patient based restenosis rate in such circumstances is the sum of the restenosis rate of each lesion treated. A “low restenosis” stent might, therefore, tip the balance in favour of treating multi-vessel disease with PCI rather than CABG in terms of clinical outcome. However, should all three lesions (or more) be treated with drug eluting stents (at huge expense)? Or should we adhere to our principles of “lesion based” restenosis risk? Our practice will have to be the latter, at least until costs come down. That approach also has an intrinsic “fairness”—every patient’s restenosis risk will be proportionately reduced, whether from 21% to 7% in the case of a single lesion procedure at high risk of restenosis, or from 33% to 19% in the case of a triple vessel procedure including one high and two low risk ones. On the other hand, apart from the cost of a drug eluting stent, a PCI procedure is cheaper than a CABG procedure, so some attenuation of costs for the institution may be possible if more patients are treated with PCI rather than CABG.


The data in regard to the restenosis rates quoted in table 1 are, of necessity, based upon studies performed in the mid 1990s, mostly using the Palmaz-Schatz stent and dated implantation techniques and periprocedural treatment. Nevertheless, the figures appear to be consistent with our recent, “real world”, patient based restenosis rates. Also, because no single data source contains a full spectrum of stent widths and lengths, we have had to extrapolate to construct table 1 (for example, in assuming that restenosis rates increase with stent length in the same proportion for different stent widths). It will also be appreciated that the data are derived from quantitative angiography rather than “eyeball” measurements (which most interventionists use). Equally, data from meticulously performed studies in the best centres may not apply so precisely to non-ideal lesions treated (in non-ideal ways) in everyday practice. We have, for instance, had to allow for the likelihood that the mean post-stent MLD is 10–15% (rather than 0%). Finally, the costs associated with drug eluting stents are likely to fall. When they do, their use will broaden, although the principles underlying the analysis we have employed in this paper will remain valid.


The RAVEL study included only small numbers of certain important lesion subsets; only 18% of vessels were 2.5 mm, and mean lesion length was (a rather short) 9.6 mm. There were no implantations in left main stems, bypass grafts, chronic total arterial occlusions, ostial stenoses, acute myocardial infarctions, in-stent restenoses, and bifurcations. Similar observations apply to the SIRIUS study. A “low restenosis” assumption for lesions other than the simplest may not be valid, and discretion should, therefore, be employed when considering drug eluting stent implantation in these contexts until appropriate studies have been completed. Furthermore, the Bx Velocity stent platform, upon which the Cypher stent is based, may not be suitable in certain situations—notably in tortuous vessels. It may also be the case that the development of stents eluting alternative therapeutic agents may be more appropriate in selected clinical settings. The data from these first studies with the sirolimus eluting stent are, however, so compelling that we are likely to see drug eluting stent implantation in lesions for which there is, as yet, little or no evidence for their utility (when the price comes down).


A policy of selective implantation of drug-eluting stents, in a minority of lesions most likely to benefit, seems to be a rational way to employ this new technology. If the price of the device remains as high as presently planned, however, there is a strong argument that even this extra revenue might be better spent offering PCI with conventional stents to a considerably larger number of patients. In our hospital, we are unlikely to receive sufficient extra funds to implement anything other than a highly selective implantation policy, probably employing the 15% restenosis threshold, as described in this paper. We will then audit our results to ensure that a real reduction in restenosis is achieved. We also plan to review the restenosis “threshold” for choosing a drug eluting stent that seems economically sustainable, using the data presented here, in the light of any changes in the price of the product, the savings obtained in treating restenosis, and the emergence of competitor devices.