What is the risk of stent thrombosis associated with the use of paclitaxel-eluting stents for percutaneous coronary intervention?

A meta-analysis

Literature review

We searched the MEDLINE, EMBASE, CRISP, metaRegister of Controlled Trials, and Cochrane databases for randomized clinical trials from 2001 to 2004 using the medical subject heading terms “angioplasty, transluminal, percutaneous coronary,” “stents,” “paclitaxel,” and “thrombosis.” We also hand-searched relevant journals, obtained recently presented data from national cardiology and interventional cardiology conferences, corresponded with authors and experts in the field, and used the Science Citation Index to cross-reference any articles that met our selection criteria.

Our inclusion criteria were as follows: 1) a randomized clinical trial that assigned patients to either a PES or BMS; and 2) angiographically documented thrombosis data were available for at least 30 days of follow-up. Stent thrombosis was defined as either an angiographically documented thrombosis or an event where angiographic data were not available, although the study investigators clinically presumed that a stent thrombosis occurred. Studies were excluded from analysis if both groups received DES, if sirolimus (rapamycin) was used, if paclitaxel was given orally, if other analogues of paclitaxel such as QP-2 were employed, or if reliable data could not be obtained.

We identified eight studies that compared PES to BMS (Treatment of de novo coronary disease using a single pAclitaXel elUting Stent [TAXUS]-I, -II, -IV, and -VI, ASian Paclitaxel-Eluting Stent Clinical trial [ASPECT], European evaLUaTion of paclitaxel Eluting Stents [ELUTES], DELIVER-I, and PAclitaxel-Coated logicsTENt for the CYtostatic prevention of restenosis [PATENCY]) (5–16). This represents 13 study arms as TAXUS-II had two DES arms, ASPECT had two arms, and ELUTES had four arms.

Data abstraction

Two independent reviewers abstracted the following outcomes by intention-to-treat analysis: 1) angiographically documented or presumed coronary thrombosis; and 2) length of follow-up from deployment of the index stent procedure. We also abstracted baseline information such as patient demographics as well as angiographic and procedural characteristics of the coronary artery revascularization. Discrepancies were resolved through a third reviewer. Because all eight studies were double-blind randomized clinical trials, a formal quality assessment was not used.

End points

In order to utilize data from all available studies, and thereby increase the statistical power of our meta-analysis, we considered stent-associated thrombosis at any point in time (up to one year) after the index procedure as the primary end point. In trials that examined various doses or release kinetics of paclitaxel, each DES arm was treated as a separate study for analysis. We also conducted a number of subanalyses by analyzing groups that shared various similarities. The four TAXUS trials were analyzed together because they used clopidogrel for six months in addition to aspirin, while the other four trials used a shorter duration of dual antiplatelet therapy. Similarly TAXUS-IV and -VI trials studied patients with longer lesions (mean length >12 mm), while ASPECT, ELUTES, and DELIVER-I trials examined higher doses of paclitaxel (2.7 to 3.1 μg/mm²) and were analyzed together.

Statistical methods

We tabulated the number of thromboses and persons at risk for thrombosis in each arm of a study and calculated the risk ratio (RR) (PES vs. BMS). In order to determine if the studies were similar enough to be comparable for analysis, the Cochran's Q statistic for heterogeneity was employed. To evaluate if the studies were published with a certain bias (i.e., only studies with positive results were published), the Egger's funnel plot for publication bias was constructed. We utilized the automatic “zero cell” correction so that studies with no events in a single arm would still be included for analysis. We obtained a summary estimate of the RR and 95% confidence interval (CI) for PES compared to BMS. A fixed effects model was used to calculate the summary statistic. All p values were two-tailed, with statistical significance set at 0.05. All CIs were calculated at the 95% level. All analyses were performed using STATA software v8.0. (STATA Corp., College Station, Texas).

Baseline characteristics

The eight studies that met our inclusion criteria were TAXUS-I, -II, -V, and -VI, ASPECT, ELUTES, DELIVER-I, and PATENCY (5–16). The characteristics of these studies are listed in Table 1.A total of 3,817 patients were enrolled; 1,995 were randomized to PES and 1,822 to BMS. All trials were multicentered studies. The baseline characteristics of patients randomized to PES and to BMS were similar within each study. The median ages of participants ranged from 58 to 66 years. The percentage of enrolled females ranged from 6% to 38%. Approximately one-quarter to one-third of the patients had a history of previous myocardial infarction.

Heterogeneity/publication bias

There was no evidence for heterogeneity among the studies (chi-square value for Q-statistic = 5.90 [10 degrees of freedom], p = 0.82). Similarly the p value for publication bias was nonsignificant (p = 0.86).

Risk of stent-associated thrombosis

Length of follow-up, event data, and RRs (with 95% CI) for the development of stent-associated thrombosis are listed in Table 2.Six study arms revealed a nonsignificant decrease in the risk of stent-associated thrombosis (TAXUS-IV, TAXUS-VI, ELUTES: 0.2, 0.7, and 1.4 μg paclitaxel/mm²of stent arms; and DELIVER-I), five study arms revealed a nonsignificant increase in risk (TAXUS-II SR and MR arms and ASPECT: 1.3 and 3.1 μg paclitaxel/mm²of stent arms, and ELUTES 2.7 μg paclitaxel/mm²of stent arm), and in two studies there were no events (TAXUS-I and PATENCY).

In the ASPECT trial, all cases of thrombosis occurred in individuals randomized to PES who received nontraditional antiplatelet therapy (i.e., aspirin without the use of a thienopyridine). Aspirin and cilostazol was used in these individuals for one month (n = 29) or six months (n = 8) as this was considered standard antiplatelet therapy at the participating centers. The remaining individuals in this trial received aspirin with ticlopidine (n = 120) or clopidogrel (n = 18) for one or six months. The duration of therapy for those that received clopidogrel was one month, while the duration of therapy for 78% of the individuals given ticlopidine was six months (one month of therapy for the remaining individuals on ticlopidine).

The plot of stent-associated thrombosis and the pooled RR estimate is shown in Figure 1.The hazard of stent thrombosis for PES versus BMS was not increased up to 12 months after the index stent deployment (RR = 1.06, 95% CI 0.55 to 2.04, p = 0.86).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

In trials with varying doses of paclitaxel, a = 1.3, b = 3.1, A = 0.2, B = 0.7, C = 1.4, and D = 2.7 μg paclitaxel/mm²of stent. BMS = bare-metal stent; CI = confidence interval; DES = drug-eluting stent; MR = moderate-release; SR = slow-release.

Subanalyses

When the four TAXUS trials that used aspirin and clopidogrel for six months were analyzed together, the summary RR obtained was similar to the overall effect estimate (RR = 1.01, 95% CI 0.40 to 2.53, p = 0.99). Analyzing TAXUS-IV and -VI, which enrolled patients with longer lesions (mean length >12 mm) resulted in a similar association (RR = 0.62, 95% CI 0.20 to 1.91, p = 0.41). Analyzing ASPECT, ELUTES, and DELIVER-1, which used a higher dose of paclitaxel (approximately 3.0 μg/mm²) revealed the same overall association (RR = 1.87, 95% CI 0.52 to 6.81, p = 0.34).