Bleeding Definitions Used in PCI Trials & Registries
One of the obstacles to fully understanding the impact of bleeding after PCI is the inconsistency in definitions used to define a bleeding event. A number of definitions used to scale the severity of bleeding events have been developed and the reported bleeding rate incidence has been shown to be highly dependent upon the definitions used. Steinhubl et al. analyzed bleeding data from 13 large trials evaluating antithrombotic drugs in acute coronary syndrome (ACS) in over 178,000 patients. They concluded that it is ‘undoubtedly true’ that variations in definitions used to define major bleeding have led to differences in reported rates.
Bleeding Definitions Used in Clinical Trials
Two commonly used definitions in the past were the Thrombolysis In Myocardial Infarction (TIMI) and Global Use of Strategies to Open Occluded Arteries (GUSTO) bleeding definitions ( Table 1 ).[12,13] The TIMI scale uses decreases in hemoglobin or hematocrit and intracranial hemorrhage to classify bleeding as minimal, minor or major. The GUSTO scale defines clinical events that stratify bleeding episodes into mild, moderate or severe. While some studies have used either the GUSTO or TIMI definition, others have used both, and yet others have combined selected elements of both scales. Furthermore, some studies have developed their own criteria to define bleeding events, such as in the Randomized Evaluation in PCI Linking Angiomax to Reduced Clinical Events II (REPLACE-2), the Acute Catheterization and Urgent Intervention Triage Strategy (ACUITY), the Clopidogrel in Unstable Angina to Prevent Recurrent Events (CURE), and the Safety and Efficacy of Enoxaparin in Percutaneous Coronary Intervention Patients (STEEPLE) trials.[7,8,14,15] In the 13 trials analyzed by Steinhubl, nine of them used their own definitions, other than TIMI or GUSTO.
Table 2 provides a listing of some of the bleeding definitions used in PCI clinical trials and registries. This table illustrates the wide spectrum of bleeding definitions used and demonstrates why bleeding rates can vary so widely based solely on the definitions applied. Very broad definitions will capture a larger number of bleeding events compared with very narrow definitions and will lead to a higher reported incidence of bleeding events. Even within a single clinical trial, bleeding rates may vary depending on the definitions used. For instance, in the REPLACE-2 study, bleeding episodes were reported using the TIMI and REPLACE-2 definitions. This study compared bivalirudin with provisional glycoprotein (GP)IIb/IIIa inhibitor (GPI) versus heparin with planned GPIIb/IIIa in patients undergoing elective or urgent PCI. When the TIMI definitions are used, no difference in TIMI major bleeding is noted between the two arms (0.9 vs 0.6%; p = 0.30); however, when the REPLACE-2 definitions are applied there is significantly less major bleeding in the bivalirudin group (4.1 vs 2.4%; p = < 0.001).
Another example of this is found in the STEEPLE trial, comparing two different doses of intravenous enoxaparin with unfractionated heparin (UFH) in patients undergoing PCI. Using the STEEPLE major bleeding definition, there is significantly more bleeding in the UFH group than in either enoxaparin group. If one uses the TIMI major bleeding definition, there is no difference between any of the groups. Furthermore, if one applies the GUSTO moderate or severe definitions, there is a difference in bleeding rates between the UFH group and the low-dose enoxaparin group, but no difference between the enoxaparin groups or between the UFH group and high-dose enoxaparin group.
Bleeding Definitions Used in Registries
Another confounder in delineating an accurate rate of bleeding complications comes from the variations in the source data. For example, clinical trial data that report bleeding rates are very different from registry data for several reasons. First, registry data can more accurately represent a ‘real world’ demographic of patients, and often include patients with more comorbidities who are often not candidates for clinical trials. Second, the way in which bleeding events are captured in clinical trials differs from how these events are noted in registries. In clinical trials, patients are often prospectively evaluated for bleeding events by study coordinators or clinicians, while bleeding events reported from registry data are often found through retrospective chart review. Owing to this difference in identifying bleedings, the rate of bleedings in registries are usually based on data elements that are readily identifiable from chart review such as transfusions or surgical interventions. Therefore, variation in patient populations, data capture and definitions all likely lead to differences in sensitivity of detecting bleeding events.
Kinnard et al. reported a three-hospital registry experience on 10,974 patients undergoing PCI from 1991 to 2000 on the incidence and predictors of bleeding after PCI. They used the TIMI bleeding definition and noted major bleeding in 5.4% and minor bleeding in 12.7%, with a blood transfusion given to 5.4% of patients. The majority of bleeding events in both the major and minor bleeding groups were related to vascular access-site hematomas. They also reported the independent risk factors for bleeding, after adjustment for confounders: age, procedural hypotension, intra-aortic balloon pump use, chronic renal insufficiency, systemic hypertension history, and the use of abciximab, which were all independently associated with in-hospital bleeding events.
The National Heart, Lung, and Blood Institute (NHLBI) Dynamic Registry included a cohort of 6656 patients undergoing PCI enrolled at multiple centers, of which 97% had femoral access. Access-site hematomas requiring blood transfusions occurred in 1.8% of these patients. Older age, lower BMI, female sex, history of renal, cerebrovascular, peripheral vascular or pulmonary disease, and hypertension were significantly associated with hemorrhage.
The Global Registry of Acute Coronary Events (GRACE) analyzed data from 24,045 patients with ACS and found the overall major bleeding rate to be 3.9%. Right-heart catheter and PCI were independently associated with an increased risk of bleeding. Among the patients in the registry who had PCI, female sex, advanced age and renal insufficiency were associated with increased bleeding risk. Among all major bleeds recorded in this registry, nearly a quarter (23.8%) of them were vascular access site bleeds.
Data from over 300,000 patients undergoing PCI from the National Cardiovascular Data Registry have been used to develop a risk model to predict the risk of in-hospital bleeding after PCI. Bleeding was defined as transfusion, a drop in hemoglobin greater than 3 g/dl, or an access-site hematoma greater than 10, 5 or 2 cm in the femoral, brachial or radial site, respectively ( Table 2 ). With this definition, the incidence of bleeding was relatively low (2.5%). Significant predictors of bleeding included female sex, age, renal insufficiency, prior PCI, cardiogenic shock, emergent/urgent PCI and chronic obstructive pulmonary disease.
As outlined above, the reported incidence of bleeding depends highly on several factors: the data source (clinical trial vs registry), the patients, concomitant medical and procedural strategies (e.g., antithrombotic regimens, vascular access site), and the bleeding definition used. Despite these differences, there are some consistencies across studies. In patients undergoing PCI, the primary site of bleeding is the vascular access site. Older patients, females and patients with renal insufficiency are consistently identified as being at high risk for bleeding complications. These data are valuable in determining strategies to reduce bleeding risk, as will be discussed later.
Interv Cardiol. 2009;1(1):51-62. © 2009 Future Medicine Ltd.
Cite this: Impact of Bleeding Complications on Outcomes After Percutaneous Coronary Interventions - Medscape - Oct 01, 2009.