OUP user menu

Is bare-metal stenting superior to balloon angioplasty for small vessel coronary artery disease? Evidence from a meta-analysis of randomized trials

Pierfrancesco Agostoni, Giuseppe G.L. Biondi-Zoccai, Gabriele L. Gasparini, Maurizio Anselmi, Giorgio Morando, Marco Turri, Antonio Abbate, Eugene P. McFadden, Corrado Vassanelli, Piero Zardini, Antonio Colombo, Patrick W. Serruys
DOI: http://dx.doi.org/10.1093/eurheartj/ehi116 881-889 First published online: 28 January 2005


Aims To compare, by meta-analytical techniques, the clinical impact of bare-metal stenting vs. balloon angioplasty for the treatment of lesions in small coronary arteries.

Methods and results We included trials with random allocation and prospective comparison of angioplasty vs. stenting, reference vessel diameter <3 mm, and follow-up ≥6 months. Random effect odds ratios (OR) for death, myocardial infarction (MI), repeat revascularization (RR), and major adverse cardiac events (MACEs) were computed. In a pre-specified subgroup analysis, we compared stenting with optimal (post-procedural stenosis <20%) and suboptimal (>20%) angioplasty. Thirteen studies (4383 patients) were selected. No differences were found in terms of death and MI, while MACEs, mainly driven by RR, were significantly less common after stenting (17.6%) than after angioplasty (22.7%), OR 0.71 (0.57–0.90). Heterogeneity among trials was present. When considering only optimal angioplasty, MACE rates were homogeneously similar, 17.9 vs. 21.1%, OR 0.86 (0.66–1.11). If angioplasty were suboptimal, MACEs were significantly more common after angioplasty (24%) than after stenting (17.3%), OR 0.62 (0.44–0.88).

Conclusion Stenting is superior to balloon angioplasty for the treatment of small vessels, in particular after suboptimal angioplasty. However, MACE and RR rates remain high after stenting, and the advantage of stent over angioplasty is moderate. An optimal balloon angioplasty strategy (with provisional stenting) may achieve results not inferior to routine stenting.

  • Meta-analysis
  • Stent
  • Balloon angioplasty
  • Small coronary arteries


Percutaneous transluminal coronary angioplasty (PTCA), by means of simple balloon inflation, is an established technique for the treatment of patients with symptomatic coronary artery disease.1 One of its major limitations is the occurrence of restenosis that requires repeat intervention.2 The use of stents has been shown to reduce both angiographic restenosis and the need for repeat intervention in discrete lesions and large vessels.3,4 Since then, the use of stenting has also increased radically for ‘non-STRESS/BENESTENT’ lesions, despite lack of evidence of benefit.5 Indeed, the relative merits of stenting and balloon angioplasty remain contentious in lesions in small vessels, long and diffuse lesions, bifurcation, and ostial lesions, and in-stent restenotic lesions. In particular, lesions in small coronary arteries (with a diameter of <3 mm) are the most relevant in terms of prevalence, accounting for 40–50% of all coronary stenoses.5

Several recent randomized trials have compared stenting and PTCA, in terms of clinical and angiographic outcomes, in coronary arteries with a reference vessel diameter (RVD) of <3 mm, with conflicting and overall inconclusive results. As systematic overviews may provide more precise effect estimates with greater statistical power,6 we performed a meta-analysis of all the trials directly comparing these two percutaneous coronary strategies.


Search strategy

Two trained investigators (P.A., G.B.-Z.) independently searched MEDLINE and CENTRAL for eligible studies published between January 1994 and August 2004. Search keywords included: ‘small AND coro* AND (stent* OR intervention*)’ (where * denotes a wildcard). MEDLINE was searched using the method described by Biondi-Zoccai et al.7 Conference proceedings from the 2000–2004 American College of Cardiology, American Heart Association, European Society of Cardiology, Transcatheter Cardiovascular Therapeutics scientific sessions were also searched.

Selection strategy

Citations initially selected by systematic search were first retrieved as title and/or abstract and screened independently by two reviewers (P.A., G.B.-Z.). Potentially relevant reports were then retrieved as complete manuscripts and assessed for compliance with inclusion and exclusion criteria.

Inclusion criteria for retrieved studies were: (i) prospective comparison of stenting vs. PTCA in coronary arteries with an RVD <3 mm, (ii) randomized treatment allocation, (iii) intention-to-treat analysis, (iv) follow-up of at least 6 months.

Exclusion criteria were: (i) not retrievable or unclear data, (ii) utilization of drug-eluting stents (DESs), (iii) utilization of devices different from stents (i.e. atherectomy and cutting balloon), (iv) use of anti-platelet drugs different from acetylsalicylic acid or thienopyridines.

Data abstraction and validity assessment

Two non-blinded reviewers (P.A., G.G.) independently performed data abstraction on pre-specified structure collection forms and evaluated study quality according to the Jadad score,8 allocating 1 point for the presence of each of the following: (i) study defined as randomized, (ii) study defined as blinded, (iii) clear description and discussion of withdrawals and dropouts. If randomization and blinding were appropriate, one additional point was added for each, otherwise a point was deducted. Thus, the total score ranged from 0 to 5. Divergences in data abstraction and quality evaluation were resolved by consensus.

Study characteristics

The primary endpoint was the incidence of major adverse cardiovascular events (MACEs), defined as the composite endpoint of death, myocardial infarction (MI), and repeat revascularization (RR), evaluated at the longest available follow-up for each trial. More specifically, all-cause death; MI, defined as recurrent ischaemic chest pain associated with new electrocardiographic changes or elevation of cardiac enzymes (according to each study); and RR, defined as target lesion or vessel revascularization, by means of repeated percutaneous coronary intervention (PCI) or coronary artery bypass graft (CABG) surgery, were also assessed. All the outcomes analysed were used as defined in individual trials.

Crossover from PTCA to stent was allowed only as bail-out, in case of abrupt or threatened closure of the vessel or in the case of residual stenosis >50%. Crossover from stent to PTCA was defined as failure to implant the stent. However, the outcomes were assessed on an intention-to-treat basis.

In the light of some expert opinion that stenting and optimal PTCA were equivalent strategies in small vessels,9,10 we also performed a pre-defined subgroup analysis dividing the studies according to the mean post-procedural diameter stenosis (DS) of the PTCA. A PTCA was considered optimal when a residual DS was <20%. We, thus, divided the studies into two groups: those with a mean post-procedural DS (in the PTCA group) of <20% or >20%.

Secondary endpoints included all the angiographic data. In particular, we evaluated: angiographic restenosis, defined as DS>50% at follow-up coronary angiography, performed 6 months after enrolment; baseline RVD; minimal luminal diameter (MLD) and DS before intervention, immediately after, and at follow-up; acute gain and late loss.

Data analysis and synthesis

Statistical analysis was performed using the Review Manager 4.2 freeware package.11 Dichotomous variables are reported as proportions and percentages, continuous variables as mean ± standard deviation. Binary outcomes from individual studies were combined with both the Mantel–Haenzel fixed effect model and the DerSimonian and Laird random effect model (the latter is used to present the data because it is more robust and conservative than the former and less subject to error if heterogeneity is present),12 whereas continuous variables were compared using the DerSimonian and Laird random effect model.12 Odds ratios (ORs) with 95% confidence interval (CI) and weighted mean differences (WMDs) with 95% CI were used as summary statistics for the comparison of, respectively, dichotomous and continuous variables. Reported values were two-tailed and results were considered statistically significant at P<0.05. Formal Cochran Q χ2 tests were performed to investigate heterogeneity between trials (respective scores, degrees of freedom, and P-values are reported) and statistical heterogeneity was considered substantial if the Cochran Q test yielded a P-value of <0.10.12

In order to assess the sensitivity of the results, several subgroup analyses were performed. In addition to the pre-specified optimal vs. suboptimal PTCA analysis, we also evaluated only papers published in extenso, high-quality studies, trials evaluating smaller vessels (we analysed trials including vessels with a maximum RVD<2.8 mm), trials including coated or bare-metal stents, and trials evaluating stents specifically designed for small vessels.

Computation of power and number-needed-to-treat (NNT, with 95% CI) extrapolated from pooled random effect risk differences was finally made in order to explicitly assess limitations and robustness of the systematic overview.12

A funnel plot of treatment effect vs. study precision was created for the primary outcome, and the Egger test was computed to look for possible publication bias.12


Search results and study selection

Overall, 484 citations were retrieved. Several reports were excluded at the title/abstract level because they were not pertinent. Finally, 62 eligible citations were assessed for compliance with inclusion and exclusion criteria: 52 studies were excluded because of observational evaluation, duplicate publication, non-randomized treatment allocation, post hoc analysis of randomized trials, or utilization of devices different from stents. Two trials were excluded because DESs were used.13,14 Eight published studies were finally selected.1522 Additional search of conference proceedings led us to find five pertinent randomized trials, published as abstracts, fulfilling our inclusion criteria,2327 and two potentially eligible studies, both excluded, one because it was not completed and the data were not clearly retrievable28 and one because cilostazol was used instead of acetylsalicylic acid.29 Unpublished studies were included in the analysis because of the importance of the so-called ‘grey’ literature.30

Data abstraction of the 13 studies selected was then performed and individual researchers contacted in cases of incomplete reporting, thus obtaining all the relevant information for each trial (in particular, the methods and results of four unpublished trials23,2527 were accessed at http://www.tctmd.com) (Table 1).

View this table:
Table 1

Description of included studies

StudyYears of enrolmentLocationFollow-up (months)Quality
Park et al.151997–98South Korea163
SISA1997–99Canada, Europe, Asia, Brazil122
SISCA1998–99Norway, Sweden123
Kinsara et al.212003aSaudi Arabia61
ISAR-SMART II2000–02Germany121
LASMAL2001South America61
LASMAL II2001–03South America121

aYear of publication.

Patients' baseline characteristics

The 13 studies included in the final analysis randomized 4383 patients, 2097 to PTCA and 2286 to stenting. The mean age was 62.1±10 years. On average, males accounted for 72.7% of subjects and 28.9% of patients suffered from diabetes. The indication for PCI was unstable angina in 31.5% of patients. All patients treated with stenting received double antiplatelet therapy (acetylsalicylic acid and a thienopyridine) for 1 month. In most studies, PTCA patients received only acetylsalicylic acid, apart from one16 in which they were also treated with ticlopidine for 2 weeks and another27 in which clopidogrel was also given for 1 month. Extensive use (>50%) of abciximab was present in only two studies,16,27 in two others25,26 glycoprotein IIB/IIIA inhibitors were used, respectively, in 22 and 40% of patients, while in the remaining trials their use ranged from 0 to 5%. Overall, there was no major imbalance regarding the use of glycoprotein IIB/IIIA inhibitors in the two groups (Table 2). A bare-metal stent was used in six trials while a stent coated with heparin or silicon carbide or phosphorylcholine was used in the remainder (see Table 3 for details).

View this table:
Table 2

Clinical characteristics of the patients of the included studies

StudyNo. of patientsAge (years)aMale (%)Diabetes (%)Unstable angina (%)Anti-thrombotic therapy
Park et al.1512061±8631219ASA, TP
ISAR-SMART40466±11772540ASA, TP, GP (100%)
BESMART38161±10761747ASA, TP
SISA35160±10671932ASA, TP, GP (5%)
SISCA14563±11651323ASA, TP, GP (1%)
COAST58861±10741916ASA, TP
Kinsara et al.2120255±11745615ASA, TP
SVS49661±10671624ASA, TP
ISAR-SMART II50266±107328NAASA, TP, GP (50%)
LASMAL246NA772662ASA, TP, GP (22%)
LASMAL II220NA7510062ASA, TP, GP (40%)

ASA, acetylsalicylic acid; GP, glycoprotein IIB/IIIA inhibitors; NA, not available; TP, thienopyridines (ticlopidine or clopidogrel).

aData are mean ± SD.

View this table:
Table 3

Angiographic characteristics of the included studies

StudyBaseline RVD (mm)Lesion length (mm)aPost-PTCA DS (%)aCrossover PTCA→stent (%)Crossover stent→PTCA (%)Type of stent
Park et al.15<3.0<1514±13200Bare-metal
SISCA2.1–3.011.3±4.425±1314.14.1H coated
COAST2.0–2.67.6±3.312±1627.21.3H coated, Bare
Kinsara et al.21≤2.511.4±6.224±915.11H coated
SVS2.0–3.09.4±3.325±1028.45.6SC coated
ISAR-SMART II≤2.511.2±6.218±1740.23.6PC coated
LASMAL2.0–2.910.1±711±6180PC coated
LASMAL II2.0–2.99.7±711±623.80PC coated

DS, diameter stenosis; H, heparin; PC, phosphorylcholine; SC, silicon carbide.

aData are mean ± SD.

Baseline angiographic data

RVD was 2.33±0.29 mm in the stent group and 2.31±0.29 mm in PTCA patients, with a statistically significant WMD of 0.02 mm (95% CI 0.00–0.04) (P=0.02). Mean lesion length was 10.1±5.2 mm (Table 3). DS and MLD pre-procedure were comparable in both groups; respectively, 68±13.4% and 0.73±0.33 mm in the stent group and 67.5±13.1% and 0.73±0.32 mm in the PTCA group. The overall rate of crossover was 22.2% (466/2097) among PTCA patients, while it was 2.4% (55/2286) in the stent group (see Table 3 for details). Post-procedural mean DS and MLD were significantly better after stenting than after PTCA; respectively, 11.6±10.1 vs. 21.5±12.2% and 2.16±0.37 vs. 1.86±0.37 mm (P<0.001 for both). WMD for DS was 9.4% (7.0–11.7), while for MLD it was 0.29 mm (0.22–0.36). Accordingly, acute gain was 1.41±0.44 mm after stenting vs. 1.09±0.46 mm after PTCA, with a statistically significant WMD of 0.34 mm (0.26–0.42) (P<0.001).

Clinical outcomes

The median follow-up time was 8 months (inter-quartile range 6–12). Death and MI rates did not differ significantly between groups; respectively, 30/2286 (1.3%) and 72/2286 (3.1%) in the stent group vs. 37/2097 (1.7%) and 89/2097 (4.2%) after PTCA. In the pre-hoc specified subgroups based on an optimal PTCA result, these results were comparable to those seen in the overall population (Figure 1A and B).

Figure 1 Comparison of the risk of death (A), MI (B), RR (C), and MACEs (D) in patients treated with stenting vs. PTCA in each study, in the two pooled subgroups according to post-PTCA DS, and in the overall population, showing OR and 95% CI.

Stenting showed an overall significant reduction of the risk of RR when compared with PTCA (P=0.02); respectively, 341/2286 (14.9%) vs. 393/2097 (18.7%), with an OR of 0.76 (0.61–0.95). However, there was significant heterogeneity among the trials (P=0.04). According to the pre-hoc specified subgroup analysis, all the studies in which an optimal PTCA was performed [mean post-procedural DS: 15% (range 11–19%)], were homogeneous (P=0.62) and no difference was noted between stent and PTCA in terms of RR (P=0.54); respectively, 172/1145 (15%) vs. 156/935 (16.7%), with an OR of 0.93 (0.73–1.18). In contrast, among trials in which a suboptimal PTCA result was achieved [mean DS: 27% (24–33%)], stenting yielded a significant reduction of RR vs. PTCA (P=0.02); respectively, 169/1141 (14.8%) vs. 237/1162 (20.4%), resulting in an OR of 0.64 (0.45–0.92) (Figure 1C).

MACE rates were driven mainly by RR differences (Figure 1D). They were significantly less common among stent than among PTCA patients (P=0.004); respectively, 402/2286 (17.6%) vs. 477/2097 (22.7%), OR 0.71 (0.57–0.90). Heterogeneity among trials was also present (P=0.01). However, when considering only optimal PTCA, results were homogeneous (P=0.27) and MACEs were similar (P=0.24), 205/1145 (17.9%) in the stent group vs. 198/935 (21.1%) the PTCA group, OR 0.86 (0.66–1.11). If PTCA was suboptimal, MACEs were more common in PTCA [279/1162 (24%)] than in stent patients [197/1141 (17.3%)], OR 0.62 (0.44–0.88) (P=0.007).

Angiographic follow-up analysis

Of 3957 patients, 3241 (81.9%) underwent follow-up coronary angiography in the 12 studies that reported this result. The RAP investigators23 did not provide this number, but only the proportion of patients with restenosis. Follow-up DS was significantly less severe after stenting than after PTCA; respectively, 39.5±23.1 vs. 43.2±21.7% [WMD 3.1% (0.1–6.1), P=0.04], while MLD was larger; respectively, 1.41±0.56 vs. 1.31±0.52 mm [WMD 0.09 mm (0.03–0.16), P=0.006]. However, late loss was more evident after stenting (0.74±0.57 mm) than after PTCA (0.57±0.54 mm), with a significant WMD of 0.18 mm (0.10–0.26) (P<0.001). Angiographic restenosis was evident in 478/1720 (27.8%) patients in the stent group vs. 545/1521 (35.8%) in the PTCA group, yielding a significant difference between the two PCIs [OR 0.67 (0.52–0.87), P=0.003], but heterogeneity was also present in this case among all the trials (P=0.001). The subgroup analysis showed that optimal PTCA had a restenosis rate of 34.8% (265/760 patients) vs. 31.1% (294/945) for stenting, without a statistically significant difference [OR 0.84 (0.63–1.12), P=0.25]. In contrast, restenosis in the suboptimal PTCA group was 36.8 (280/761) vs. 23.7% (184/775) in the stent group; this difference was highly statistically significant [OR 0.53 (0.37–0.76), P<0.001].

Sensitivity analyses

Additional subgroup analysis is shown in Table 4. Substantial heterogeneity was present in all the subgroups and, in all these subgroups the pre-specified analysis on optimal PTCA yielded results homogeneously similar to stenting. Power calculation showed that this meta-analysis had a 1-beta error >0.99 to detect a 5.1% absolute reduction in the risk of MACE, assuming a 22.7% rate in the PTCA group and a two-tailed alpha error of 0.05. NNT calculation showed that assigning 20 (11–50) patients with a small vessel lesion to a stenting procedure is sufficient to prevent one MACE with respect to PTCA. Stent use, with respect to a PTCA with a suboptimal result, further improves to 12 (8–50) the rate of NNT to prevent a MACE. The funnel plot of effect size for primary outcome of MACE vs. study precision did not show any major imbalance with respect to possible publication bias (Figure 2), and the Egger test, to assess this bias, was not significant [beta:−1.6 (95% CI −4.3–1.1), P=0.22].

Figure 2 Funnel plot of effect size for primary outcome of MACEs vs. study precision including all the studies.

View this table:
Table 4

Subgroup analysis

Subgroups (no. of studies)No. of patientsMACE rate (%)OR (95% CI)Heterogeneity P-value
High-quality papers (3)66916.918.10.68 (0.26–1.78)0.04
Published studies (8)268717.221.60.75 (0.54–1.04)0.02
Smaller vessels (5)212218.122.90.78 (0.54–1.11)0.03
Coated stents (7)a220319.522.80.76 (0.56–1.04)0.06
Bare-metal stents (7)a237515.721.50.67 (0.50–0.91)0.08
Stents specific for  small vessels (8)276719.624.90.69 (0.51–0.94)0.009

aThe COAST trial had two stent arms, one heparin-coated, one bare-metal.


The present systematic overview of 13 randomized trials comparing elective stenting with PTCA with bail-out stenting in more than 4000 patients with small vessel coronary artery disease shows that overall stenting is safe and significantly reduces the rate of angiographic restenosis and of repeat revascularization, consequently leading to a clinically and statistically significant reduction in MACEs. However, the finding of significant heterogeneity in the overall analysis casts a light of caution on comprehensive pooled effect estimates. A reasonable explanation for this heterogeneity is suggested by the pre-specified subgroup analysis after stratification based on optimal vs. suboptimal PTCA. This analysis shows that the rates of angiographic restenosis and revascularization, in trials in which an optimal result was obtained after PTCA, are homogeneously similar to those after stenting, thus leading to similar rates of MACEs. Indeed, the overall results of our meta-analysis were influenced, to a great extent, by the differences in MACE rates between the optimal and suboptimal PTCA groups. The MACE rate was 21.1% after optimal PTCA compared with 24% when the result was suboptimal; whereas after stenting, MACEs remained stable at around 17.6%. Other indirect support of our conclusion is provided by the sensitivity analysis of subgroups: neither in the three high-quality studies nor in the trials published as full papers was stenting superior to PTCA.

Comparison with current data

A recently published meta-analysis on the same topic showed a clear advantage of stent implantation over PTCA in terms of angiographic restenosis, with a benefit which seemed more evident as the RVD became smaller.31 However, several methodological flaws in the paper call the soundness of some of the conclusions into question. Indeed, heterogeneity among studies, although significant, was not adequately assessed.12 Furthermore, non-weighted linear regression, instead of variance-weighted meta-regression, was used to test the association between RVD and the benefit of stenting.32 In fact, using the more appropriate variance-weighted meta-regression technique, no significant relationship was found.33 Finally, angiographic restenosis was the primary endpoint, thus losing the clinical impact of stenting.

Clinical implications

Stenting has been proved to be superior to PTCA in a subset of de novo lesions in coronary arteries ≥3 mm, and is associated with a 25–30% reduction of restenosis.3,4 However, some concerns were raised regarding the equivalence, in terms of restenosis, of optimal or ‘stent-like’ PTCA in comparison with stenting in de novo native lesions, as some studies showed substantial similarity among the two types of intervention34 while others provided evidence that systematic stenting was a superior strategy.35 All the aforementioned studies clearly showed that stenting yields a larger post-procedural MLD than PTCA, but also a greater late loss at follow-up control angiography. Indeed, restenosis after PTCA is determined by three factors: early elastic recoil, arterial remodelling after dilatation, and neo-intimal hyperplasia.36 Stents substantially eliminate recoil and remodelling through their mechanical effect. However, metal struts produce a greater response in terms of neo-intimal proliferation, thus leading to higher late loss than PTCA.37 The clinical advantages of stenting are related to the larger initial acute gain, which is greater than the absolute late loss, thus resulting in a relative net gain at follow-up. In fact, in small vessels, absolute late loss after stenting is equal to that in large vessels, therefore, as initial acute gain in smaller vessels is lower, the net effect is a higher restenosis rate.38 PTCA, after achieving an adequate post-procedural lumen, produces a less traumatic arterial injury than stenting, thus reducing the neo-intimal response and, consequently, the extent of late loss.39

Anyhow, it should also be emphasized that in our meta-analysis, which was on an intention-to-treat basis, ∼22% of patients randomized to PTCA received a stent as bail-out. Thus, acute complications due to abrupt vessel closure after PTCA (MI or even death) were almost eliminated with bail-out stenting, and this may explain the similar results obtained by the two PCIs in terms of ‘hard’ endpoints (death or MI). However, a strategy of optimal PTCA with bail-out stenting proved not to be inferior to systematic stenting in small vessels, underlying the possibility of using stents only in selected cases40 as bail-out or, according to our analysis, as provisional when the residual DS is >20%. This type of strategy becomes particularly useful in some subgroups of patients, such as those with contraindications to or presumed poor compliance with double antiplatelet therapy, allergy to acetylsalicylic acid, or allergy to thienopyridines.

Future treatment strategies

In any case, the rates of MACEs, repeat PCI, and restenosis remain particularly high after stenting small vessels, exceeding 17, 15, and 27%, respectively. These data are homogeneous among the entire population and all the subgroups. Thus, our analysis indirectly underlines the need for devices with improved efficacy in these lesions. While in large vessels (with an RVD >3 mm) the rates of RR and restenosis are nowadays <10%,41,42 small vessel treatment does not attain similar results. Thus, in light of the results of DES trials,13,14 these new devices become very attractive for the treatment of vessels with RVD <3 mm. However, due to economic concerns regarding the widespread utilization of DES, it should be interesting to compare, mainly from the point-of-view of cost-effectiveness, systematic DES implantation with an optimal PTCA strategy with provisional stenting, in small vessels.

Limitations of the present study

Limitations of meta-analyses are well known.6,12 Of relevance to this study, substantial statistical heterogeneity, as shown by the overall analysis, is considered by some authors a contraindication to quantitative pooling.43 However, other investigators recommend that, pending further trials exploring differences between studies, clinicians should look to a summary measure from the available trials for the best estimate of the impact of an intervention.44


Stenting is superior to PTCA for the treatment of small coronary artery lesions, in particular when a suboptimal post-PTCA result is achieved. Our data suggest that the benefits of stenting may be dependent on the relatively more common inability of PTCA alone to achieve optimal angiographic result (residual DS <20%). Indeed, a strategy of optimal PTCA (with provisional stenting) may achieve results not inferior to systematic stenting.


This study is part of an ongoing training project of the Center for Overview, Meta-analysis and Evidence-based medicine Training (COMET), based in Verona, Italy.


View Abstract