Bleeding in Patients Undergoing Percutaneous Coronary Interventions: A Risk Model From 302,152 Patients in the NCDR.

1
Bleeding in Patients Undergoing Percutaneous
Coronary Interventions A Risk Model From 302,152
Patients in the NCDR.

• Sameer K. Mehta MD, Andrew D. Frutkin MD, Sunil
  V. Rao MD, FangShu Ou MS, Eric D. Peterson MD,
  MPH, John A. Spertus MD, MPH, Steven P. Marso MD,
  on behalf of the NCDR
• The Mid America Heart Institute and Duke Clinical
  Research Institute.

American Heart Association Scientific Sessions
November 5, 2007
2
Presenter Disclosure Information

• Sameer K. Mehta MD et al.
• Bleeding in Patients Undergoing Percutaneous
  Coronary
• Interventions A Predictive Model From 302,152
  Patients in
• the NCDR.
• No authors report any disclosures or conflicts of
  interest.

3
Bleeding and PCI

• The incidence of major bleeding in patients
  undergoing PCI is as high as 5.
• Bleeding events have been associated with an
  increased incidence of adverse outcomes,
  including death, MI, and stroke.
• A tool that helps physicians assess the risk of
  bleeding in PCI patients may allow physicians to
  individualize antithrombotic and antiplatelet
  therapy while mitigating the risk of bleeding.

Goodman et al. Circulation 2003. Eikelboom et al.
Circulation 2007. Rao et al. American Journal of
Cardiology 2005.
4
Aims

• To determine the risk factors of bleeding in a
  large, contemporary, real-world cohort of
  patients undergoing PCI.
• To develop a risk model for bleeding.
• To test the validity of this risk model in
  clinically important subgroups.

5
Methods

• Version 3.04 of NCDR Cath-PCI Registry
• Contains data from PCI procedures performed from
  Jan. 1, 2004 to March 31, 2006 at over 600 U.S.
  hospitals.
• Exclusions
• - Non-index PCI
• - Patients who died same day as PCI
• - Patients with missing bleeding data
• - Centers that did not report any bleeding
  events

6
Methods (2)

• Final study population
• 302,152 patients
• 302,152 PCI procedures
• 440 U.S. hospitals
• Determined predictors of bleeding with
  generalized estimating equation models.
• Models adjusted for age, sex, weight, GFR, PCI
  status, presence of ACS, cardiogenic shock,
  intra-aortic balloon pump treatment, history of
  MI, diabetes, cerebrovascular disease, peripheral
  vascular disease, hypertension, COPD, prior PCI,
  NYHA class, prior valve surgery, Caucasian, prior
  CHF, smoker, family history of CAD, EF,
  dyslipidemia, prior CABG, and CHF.

7
Methods (3)

• Risk model training set.
• 241,512 patients (80 of total)
• Risk model validation set
• 60,640 patients (20 of total)
• Variable selection via backward selection and
  clinical judgment
• Goodness of fit determined by calibration plot
• Discrimination assessed by c-statistic
• Risk model tested in various clinically
  meaningful subgroups

8
NCDR Bleeding Definitions

• Primary Endpoint Bleeding from any source
• Percutaneous entry site
• during hospitalization
• transfusion and/or cause a drop in hemoglobin
  gt3.0 g/dl
• hematoma
• gt10cm for femoral access,
• gt2cm for radial access,
• or gt5cm for brachial access.
• Retroperitoneal
• transfusion and/or cause a drop in hemoglobin
  gt3.0 g/dl.

9
NCDR Bleeding Definitions (2)

• GI
• transfusion and/or cause a drop in hemoglobin
  gt3.0 g/dl.
• GU
• transfusion and/or cause a drop in hemoglobin
  gt3.0 g/dl.
• Other/Unknown
• During hospitalization
• transfusion and/or cause a drop in hemoglobin
  gt3.0 g/dl.

10
Select Baseline Characteristics
Age (years, median) 64
Female () 34
Caucasian () 87
Weight lt50 kg () 1.5
GFR lt30 () 10
Hypertension () 25
ACS () 66
Cardiogenic Shock () 2.0
IABP () 2.2
2b/3a Use () 48
Direct Thrombin Inhibitor () 32
11
Results

• Incidence of Bleeding in Training Set
• 2.5

12
Risk Factors for Bleeding- Adjusted Analysis
Variable OR 95 CI ? Square
Female 1.74 1.64-1.85 319.3
Age (per 10 yrs) 1.36 1.31-1.42 223.6
GFR (per 10 ml/min decrease) 1.11 1.10-1.13 210.2
Prior PCI 0.69 0.64-0.73 128.9
Cardiogenic Shock 1.87 1.66-2.10 104.7
Emergent/Salvage PCI Urgent PCI 2.22 1.46 1.98-2.49 1.34-1.57 81.7
COPD 1.31 1.23-1.39 70.6
All p values lt0.001
13
Risk Factors for Bleeding- Adjusted Analysis
Variable OR 95 CI ? Square
NYHA Class 3 NYHA Class 4 1.14 1.42 1.05-1.24 1.30-1.56 41.4
Non-STEMI/ Unstable Angina STEMI 1.13 1.49 1.04-1.22 1.32-1.69 35.9
Prior Valve Surgery 1.61 1.34-1.94 24.8
CVD 1.16 1.09-1.24 19.1
Intra-aortic balloon pump 1.95 1.41-2.70 16.1
PVD 1.15 1.07-1.23 13.6
HTN 1.12 1.05-1.19 12.5
Weight (per 5 kg decrease) 1.02 1.01-1.03 11.9
14
Overall Model
Predicted Bleeding ()
N 60,640 C Statistic 0.73
Observed Bleeding ()
15
Subgroup Analyses
Predicted Bleeding ()
N 30,872 C Statistic 0.72
N 9,130 C Statistic 0.70
Observed Bleeding ()
16
Elective PCI Patients
Predicted Bleeding ()
N 29,733 C Statistic 0.67
Observed Bleeding ()
17
Antithrombotic Therapy
Unfractionated Heparin plus IIb/IIIa
Unfractionated Heparin
Predicted Bleeding ()
N 22,666 C Statistic 0.72
N24,969 C Statistic 0.73
Observed Bleeding ()
18
Antithrombotic Therapy
Low Molecular Weight Heparin
Direct Thrombin Inhibitors
Predicted Bleeding ()
N 10,108 C Statistic 0.68
N 19,316 C Statistic 0.73
Observed Bleeding ()
19
Conclusions

• Identified risk factors for bleeding in PCI
  patients
• Developed a risk model that predicted the risk of
  bleeding in patients undergoing PCI
• Model performed well in various clinically
  important subgroups

20
Limitations

• Bleeding definitions differ between NCDR and
  TIMI/ GUSTO
• Low reported incidence of bleeding events

21
Future Directions

• Risk Score for Bleeding.
• Implementation of predictive model/ risk score
  into randomized studies of patients undergoing
  PCI.