Arterial Puncture Closing Devices Compared With Standard Manual Compression After Cardiac Catheterization:  Systematic Review and Meta-analysis FREE

The number of annual coronary interventions exceeds 500 000 in the United States^1- 2 and is estimated to exceed 1 million worldwide.^1,3 Along with major complications, such as coronary artery dissection, thrombus formation, and coronary artery spasm leading to acute occlusion of a coronary vessel, there are additional complications related to the site of peripheral arterial access, including hematoma, bleeding, arteriovenous fistula, and pseudoaneurysm. The reported overall vascular complication rates range from 1.5% to 9%,^4- 8 and 20% to 40% of patients who experience such complications require surgical repair.^7- 8

After removal of the catheter sheath, hemostasis is usually achieved by manual compression at the vascular access site with or without the use of adjunctive mechanical compression devices. Thereafter, prolonged bed rest is often recommended. Bed rest and compression are associated with discomfort to the patient and may have cost implications. Arterial puncture closing devices (APCDs) have been developed to avoid manual compression and shorten bed rest.

Various techniques to replace manual compression have been designed: collagen plugs with or without an anchor from inside the artery (Angioseal,^9- 20 Vasoseal^21- 30); balloon-positioning catheters combined with bovine microfibrillar collagen and thrombin (Duett^31- 32); a suturelike stitch placed around the femoral artery (Perclose^15,33- 38 [including Techstar and Prostar]). We estimate that APCDs are now being used in 50% of patients undergoing a percutaneous coronary intervention.

We performed a systematic review of the literature to assess the safety and efficacy of APCDs compared with manual compression in patients after coronary angiography or percutaneous vascular interventions.

We searched MEDLINE (1966-January 2003), EMBASE (1989-January 2003), PASCAL (1996-January 2003), BIOSIS (1990-January 2003), and CINHAL (1982-January 2003) databases, and the Cochrane Central Register of Controlled Trials. We searched for terms describing the intervention with and without a filter to identify randomized controlled trials. The detailed search terms can be provided by the authors on request. We also contacted companies distributing such devices (Euromed, Intramed, St Jude, and Crosstec; all located in Vienna, Austria) and asked whether they were aware of any published or unpublished trials on APCDs. We also sent the results of our search to specialists in the field and asked if they were aware of any published or unpublished trials not identified by our search. We also searched references of all included articles for potentially relevant trials. We did not apply language restrictions.

We included randomized trials that compared APCDs with exclusive manual compression or standard manual compression of the femoral artery. Standard manual compression was defined as compression with the facultative use of adjunctive mechanical compression devices. Mechanical devices use a stand with a compression disk (eg, C-clamp) or a compression arch with a pneumatic dome (Femostop). For reasons of simplicity, we will refer to both manual compression and standard manual compression as standard compression.

We included all articles irrespective of publication length; that is we did not exclude articles published as abstracts or short reports, even though critical appraisal of such publications is limited.

Primarily we were interested in safety data. Before extracting data, we prospectively defined the incidence of the following vascular complications: hematoma at the puncture site; bleeding at the puncture site; formation of an arteriovenous fistula at the puncture site; and formation of a pseudoaneurysm at the puncture site. Further, we extracted other end points as reported by the authors (such as surgical intervention at the puncture site, receiving a blood transfusion, leg ischemia). For each end point, we used the definition given by the study authors.

Efficacy data (time to hemostasis, time to ambulation [ie, duration of bed rest], and time to discharge from the hospital) were also defined in advance.

Two reviewers abstracted the data independently and in duplicate to a predefined form. The results were compared and disagreements were resolved by discussion among at least 3 reviewers. We recorded whether the trial was reported according to the CONSORT criteria.³⁹ We considered the most important criteria to be adequacy of allocation concealment (yes vs unclear or no), whether the analysis was according to the intention-to-treat principle (explicitly reported vs not), and if the person assessing the outcome was blinded to the intervention group.⁴⁰ We calculated a simple score, ranging from zero (none of the 3 quality items fulfilled) to 3 (all items fulfilled). This score provides an assessment of study quality but was not used for sensitivity analysis.

The safety end points were binary and we calculated risk ratios and 95% confidence intervals (CIs) for each trial. When no outcomes occurred in 1 or both groups, 0.5 was added to each cell of the respective contingency table. When trials reported ordinal categories (eg, increasing hematoma size), we collapsed the categories to avoid overdispersion. The efficacy end points were continuous. We excluded trials with continuous end points from the analysis when the SD, SE, or 95% CIs were not reported, or when the point estimate was reported as a median only. We used random-effects models to combine the risk ratios and continuous outcomes. The effect was combined over all devices. We used STATA statistical software (Release 8, STATA Corp, College Station, Tex) for all analyses.

We used random-effects meta-regression models to assess whether predefined clinical variables influence effect size. The following variables were selected: manual compression control group vs standard manual compression control group; type of device; and diagnostic angiography vs percutaneous coronary intervention. Furthermore, we assessed the impact of study quality on the effect size: allocation concealment (reported vs not reported); blinded outcome assessment (reported vs not reported); and intention-to-treat analysis (reported vs not reported).

We computed the Cochrane Q by summing the squared deviations of each study's estimate from the overall meta-analytic estimate, weighting each study's contribution in the same manner as in the meta-analysis. We used the Q together with the resulting degrees of freedom (df) to calculate the proportion of variation due to heterogeneity (I²)[ = (Q − df)/Q].⁴¹ I² quantifies the degree of heterogeneity. The advantage is that it may be calculated and compared across meta-analyses of different sizes, different types of study, and using different types of outcome data. Funnel plots were drawn to assess visually whether there was evidence of publication bias. We also used a regression method to quantify publication bias or heterogeneity.⁴²

The electronic search resulted in 1797 hits. We considered 117 articles to be potentially eligible and retrieved the full-text versions (Figure 1). We contacted specialists in the field and manufacturers of the APCDs and found another 8 articles. Seven potentially eligible articles were identified by screening the references of the articles selected. Our search was intended to be highly sensitive at the cost of specificity. We scrutinized these articles (n = 132) more closely and excluded another 102 for the following reasons: duplicates (n = 23), different study aim (n = 8), not containing original data (n = 1), not a randomized controlled trial (n = 52), and not comparing APCDs with standard compression (n = 18). We finally included 30 studies in which APCDs were compared with standard compression. Several were primarily published as abstracts and later as full articles in peer-reviewed journals; in these cases we used the full article version. Some authors published several studies on the same topic. If an overlap of data was suspected, authors were contacted and asked for clarification. When authors did not reply, the most recent and most detailed publication was included. After excluding duplicates (n = 20), 56 articles were potentially eligible for analysis. We finally included 30 studies in which APCDs were compared with standard compression (Table 1 and Figure 1).

Generally, trial quality was fair (Table 1). Six studies^{12,16,26,31,33,35} reported allocation concealment. Four studies^14,26,29,31 explicitly reported an intention-to-treat analysis (in another 15 studies such an analysis was possible). Three studies^23,26,31 explicitly reported blinded outcome assessment. There were only 2 studies^26,31 in which allocation concealment, intention-to-treat analysis, and blinded outcome assessment were reported.

Hematoma. Nineteen publications^{9,11- 14,16,22- 26,28- 29,33- 38} with 21 comparison groups assessed the risk of hematoma at the puncture site. The definition of hematoma varied between studies: some authors simply mention the presence or absence of hematoma, others use various sizes for grading. If authors reported 2 or more categories of increasing hematoma size, the groups were collapsed. When comparing any APCD with standard compression, the relative risk (RR) of groin hematoma was 1.14 (95% CI, 0.86-1.51) (P = . 35) (Figure 2). The proportion of variation due to heterogeneity (I²) was moderately large (33%). The funnel plot showed no evidence of overt publication bias or heterogeneity (mean bias, 0.01; 95% CI, −1.07 to 1.09).

Local Bleeding. The RR of groin bleeding was 1.48 (95% CI, 0.88-2.48) in the group with APCDs compared with standard compression (P = .14) (Figure 2).^{10- 14,16,18,24,27,29- 31} The proportion of variation due to heterogeneity was 38%. In the funnel plot there was no overt evidence of publication bias but rather heterogeneity (mean bias, 1.24; 95% CI, 0.02-2.46). This was caused by a single trial in which bleeding was assessed in a meticulous way.¹⁶

Arteriovenous Fistula. The RR of arteriovenous fistula formation at the puncture site was 0.83 (95% CI, 0.23-2.94) in the group with APCDs compared with standard compression (Figure 3) (P = .77).^{24,26- 27,33- 35} The proportion of variation due to heterogeneity was 0%. The mean bias was −0.14 (95% CI, −1.34 to 1.06).

Pseudoaneurysm. The RR of developing a pseudoaneurysm at the puncture site was 1.19 (95% CI, 0.75-1.88) in the group with APCDs compared with standard compression (Figure 3) (P = .46).^{9,11- 12,14,23- 27,29- 30,33- 36,38} The proportion of variation due to heterogeneity was zero. There was no evidence of overt publication bias or heterogeneity (mean bias, 0.21; 95% CI, −0.59 to 0.90).

Other End Points. We assessed several other end points that were reported by authors, but that we did not define a priori. The RR of a surgical intervention at the puncture site was 1.61 (95% CI, 0.83-3.14) in the group with APCDs (15 comparisons, 1991 patients) compared with standard compression (1675 patients).^{13- 14,18,26- 31,33- 35,38} The RR of receiving a blood transfusion was 1.21 (95% CI, 0.57-2.55) in the group with APCDs (14 comparisons, 1737 patients) compared with standard compression (1424 patients).^{13- 14,26- 29,31,33,35,37- 38} The RR of arterial leg ischemia was 2.10 (95% CI, 0.97-4.58) in the group with APCDs (7 comparisons, 1036 patients) compared with standard compression (1060 patients).^{20,23,26- 27,33- 35}

Time to hemostasis was shorter in the group with APCDs compared with standard compression (mean difference, 17 minutes; range, 14-19 minutes).^{10,12- 13,16,18,21,24- 25,28- 30,33- 37} There was a high degree of statistical heterogeneity (I²= 96%), which may partly be explained by clinical heterogeneity; time points of assessment were often driven by protocol. Selective nonreporting of smaller differences (that is, not strongly favoring APCDs) may be possible (mean bias, −7.56; 95% CI, −10.87 to −4.26) (Figure 4).

Duration of bed rest was also shorter in the group with APCDs (1283 patients) compared with standard compression (14 comparisons, 1142 patients) (mean difference, 10.8 hours; 95% CI, 8.5-13.1 hours).^{10,12,18,21,28- 29,33- 36} There was strong evidence of statistical heterogeneity (I² = 98%) or publication bias in favor of APCDs (mean bias, −6.08; 95% CI, −11.44 to −0.72). However, the data are difficult to interpret because this end point was mainly driven by protocol and not necessarily a result of the intervention.

Duration of hospital stay was shorter in the group with APCDs (4 comparisons, 447 patients) compared with standard compression (368 patients) (mean difference, 0.6 days; 95% CI, 0.1-1.1 days).^12,25,32,35 There was some evidence of statistical heterogeneity (I² = 58%) and publication bias (0.69; 95% CI, −6.23 to 7.61). This end point was mainly driven by protocol.

The association between the intervention and the outcomes was not influenced when looking at an exclusive manual compression control group vs a standard manual compression control group (P>.13 for all 4 outcomes). It was also not influenced by type of APCD (Vasoseal, Angioseal, Duett, Perclose; P>.20 for all 4 devices and all outcomes) and whether a diagnostic angiography or a coronary intervention was performed (P>.40 for all outcomes).

When the analysis was not explicitly reported to be intention-to-treat, trials found a harmful effect less often for APCDs for hematoma (RR, 0.59; 95% CI, 0.29-1.06; P = .08) and pseudoaneurysm (RR, 0.19; 95% CI, 0.04-0.91; P = .04). When allocation was not concealed, trials were less likely to find a harmful effect of APCDs in terms of bleeding compared with trials with reported concealment (RR, 0.42; 95% CI, 0.18-0.99; P = .05).

Other main outcomes were not influenced by study quality (all P values >.10). We repeated the analysis including only trials with explicit intention-to-treat analyses. The RR of hematoma was higher in the APCD group (1.89; 95% CI, 1.13-3.15) (4 comparisons, 414 patients in APCD group, 259 patients in control group); the RR of developing a pseudoaneurysm was also higher in the APCD group (5.40; 95% CI, 1.21-24.5) (same 4 comparisons as above). When analyzing only trials in which allocation concealment was explicitly reported, the RR of bleeding was comparable between the groups (0.97; 95% CI, 0.33-2.83) (3 comparisons, 730 patients in APCD group vs 575 patients in control group).

We are aware of 2 systematic reviews on the topic of hemostatic APCDs. The first is an economic evaluation that used a systematic review of MEDLINE to determine a decision analysis.⁴³ Search strategy, inclusion criteria, and exclusion criteria of this review are not clear. The authors did not look at methodological quality, except whether the trials were randomized or not. The authors included 4 randomized controlled trials and 16 nonrandomized studies. It is not clear how the data were quantitatively summarized. This review found no significant difference in the incidence of hematoma, blood transfusion, pseudoaneurysm, or arteriovenous fistula formation between intervention (APCD) and control groups.

A second review presented data from a MEDLINE review and from abstracts presented at selected scientific meetings.⁴⁴ Search strategy, inclusion criteria, and exclusion criteria are not clear. Methodological quality of the trials was not assessed and not all were randomized trials. The author of that review looked at hemostasis success rates and minor and major complication rates. The qualitative description of the end points is comprehensive and the author points out a great deal of heterogeneity involving the definition of end points. Major complications occurred more often when particular devices (collagen plugs without an anchor) were compared with standard compression (3.8% vs 1.7%).

In an observational study, Dangas et al⁴⁵ assessed the incidence of vascular complications comparing APCDs with manual compression after percutaneous coronary interventions. The use of an APCD was associated with a significantly higher rate of hematoma (9.3% vs 5.1%; P<.001), greater decrease in hematocrit (5.2% vs 2.5%; P<.001), and increased need for surgical repair (2.5% vs 1.5%; P = .003). In another observational study, Applegate et al⁵ evaluated the outcome of 4525 patients after percutaneous coronary intervention with concomitant glycoprotein IIb/IIIa inhibitor therapy and different types of closure methods. Compared with manual compression, the RR for developing a major complication (vascular death, vascular repair, major bleeding, vessel occlusion, loss of pulse) was 1.06 (95% CI, 0.45-2.48) for Angioseal and 0.76 (95% CI, 0.42-1.38) for Perclose.

The patients studied in the trials in our systematic review were generally comparable with patients who undergo cardiac catheterization. A broad range of clinical, hospital, and geographic settings increase external validity. Unfortunately, the internal validity of many trials is not clear because most did not report according to a minimum of required standards.⁴⁰

It appears that low trial quality biased the results in favor of APCDs. Even when ignoring trial quality, the precision for some estimates is not satisfactory. For instance, based on our 95% CIs, local surgery may be less often necessary by a factor of 0.83, but it also might be required 3.41 times more often when APCDs are used. Unless prospectively defined, the ideal width for a 95% CI to assume equivalence is difficult to determine, but we can give some estimates to guide the reader. The incidence of hematoma in the manual compression group ranged from zero in some studies and reached up to 35% in others.²³ Accordingly an RR of 1.51 (the upper 95% confidence limit of our meta-analysis) translates to a number needed to harm of about 6 (Table 2). Based on our results, it is possible that for every 43 patients treated with an APCD, 1 patient will need vascular surgery. According to the literature, groin hematoma is expected in 5% to 23% of patients after manual compression, pseudoaneurysm in 0.5% to 9%, and arteriovenous fistula in 0.2% to 2%.⁴ These are worst-case scenarios, but are within the statistical precision based on the available evidence.

Although, we tried to identify all relevant trials, it is possible that some unpublished studies might have been missed. The funnel plot, for example, was not ideally shaped for trials reporting time to hemostasis. It is possible that trials reporting small differences in hemostasis time between the intervention and control groups were selectively not published (Figure 4). More likely, this is a marker of clinical heterogeneity. The time points of assessment are defined by protocol and are not necessarily the result of the intervention. One might argue that presenting summary estimates in the presence of such a high degree of heterogeneity is questionable.

It should also be taken into consideration that many trials used the earliest version of such devices. Furthermore, a learning curve for the interventionalist using such devices can be assumed. Safety may have improved because the devices have been improved and interventionalists have gained experience. However, we do not know if this translates into a clinically appreciable effect.

Some may argue that it is inappropriate to combine results for all APCDs because they are very different. In meta-regression analysis the effect was not influenced by type of device. It also was not influenced by the choice of control group (exclusively manual compression vs standard manual compression, in which mechanical devices were also used). The statistical power of such analyses is low. However, when looking at 95% CIs or P values, we do not have the impression that a large difference between groups is likely.

Finally, varying definitions for the same outcome parameters were applied.⁴⁴ For example, bleeding has not been further characterized in some studies, but has been defined in others as bleeding causing local compression,²⁷ as subcutaneous bleeding,¹¹ or as any bleeding regardless of severity.¹⁴ We assume that the meticulous definition and assessment of bleeding¹⁶ led to heterogeneity. Ideally, vascular complications should be recorded in a standardized way. Guidelines from the American College of Cardiology and the American Heart Association on percutaneous coronary intervention provide useful definitions.¹

In this systematic review, APCDs appear to be effective in terms of reducing time to hemostasis. Currently, there is not enough evidence to assess whether this translates into a clinically relevant benefit, such as reduced hospital stay. Most of the studies were of low methodological reporting quality. When looking only at studies with a higher methodological quality, complications such as hematoma and pseudoaneurysm formation occurred more often when APCDs were used.

Although APCDs are used widely in clinical practice, the safety of their use must be pursued further. It appears that manufacturers of such devices have an obligation to demonstrate efficacy and safety based on large-scale, high-quality randomized trials. Additional strategies may be found by conducting individual patient data meta-analysis, and also by tracking device use and safety data by pooling resources using registries or other relevant platforms.^46- 47