Comparative Effectiveness of Percutaneous Coronary Interventions and Coronary Artery Bypass Grafting for Coronary Artery Disease January 31, 2006 Dena Bravata, MD, MS and Mark Hlatky, MD

1
Comparative Effectiveness of Percutaneous
Coronary Interventions and Coronary Artery Bypass
Grafting for Coronary Artery DiseaseJanuary 31,
2006Dena Bravata, MD, MS and Mark Hlatky, MD
2
Todays Talk

• Background on the procedures
• Methodological issues
• Results
• Conclusions and next steps

3
What is coronary artery disease?
A narrowing of the coronary arteries that
prevents adequate blood supply to the heart
muscle. Usually caused by atherosclerosis, it may
progress to the point where the heart muscle is
damaged due to lack of blood supply. Such damage
may result in infarction, arrhythmias, and heart
failure.
4
Prevalence

• Prevalence of CAD 2004 US data 15.8 million
• Prevalence of procedures
• 2004 US data 249,000 CABGs and 664,000 PCIs

5
Coronary Artery Bypass Grafting (CABG)
6
CABG

• 1967
• Kolessov LIMA? LAD on a beating heart
• Favaloro SVG on still heart
• Procedural refinements
• arterial rather than vein grafts
• avoid the cardiopulmonary bypass machine
• smaller thoracotomy incision rather than
  sternotomy

7
Percutaneous Coronary Interventions (PCI)

• 1977 1st Coronary angioplasty by Gruntzig
• Limitation restenosis

1939-1985
8
PCI Procedural refinements Stents
Expandable metal mesh tubes that buttresses the
dilated segment, limit restenosis. Drug eluting
stents further reduce cellular proliferation in
response to the injury of dilatation.
9
Which procedure is best?
10
Research Questions

• What is the effectiveness of PCI compared with
  CABG in reducing the occurrence of adverse
  objective outcomes (stroke and death) and
  improving subjective outcomes (angina and quality
  of life) in patients with coronary artery
  disease?
• Does the comparative effectiveness of PCI and
  CABG vary based on demographics, coronary disease
  risk factors (diabetes), the extent of coronary
  artery disease, or procedure-specific
  characteristics?

11
Max
Max is a 79 yo man with well controlled HTN,
hyper-cholesterolemia, afib, and diet
controlled-DM who had a vfib arrest in the
setting of an MI. His angiogram showed proximal
vessel LAD and 3-vessel CAD. He was stabilized,
did well, and given the option of PCI or CABG.
His UCSF and Stanford doctors reviewed the
evidence with him and his family
12
Todays Talk

• Background on the procedures
• Methodological issues
• Results
• Conclusions and next steps

13
Overview of Methodologic Approach

• Our 1st MMA project slight procedural
  differences from other EPC work
• Develop a REALLY (no kidding) expert team
• Systematic literature search for RCTs
• Limited/selected searches for observational
  studies for generalizability
• Quality evaluations of primary literature
• Data extraction which time points?
• Data synthesis which outcome metric?

14
Literature Search for RCTs

• Literature searches
• 1/1966 to 8/2006
• Strategy for each source developed through
  clinician/librarian collaboration
• MEDLINE 1576 citations
• Cochrane 54 citations
• Embase 65 citations
• Additional searches (experts) 7 citations

15
Study Selection RCTs

• Procedures of Interest Any PCI technology
  (balloon angioplasty with or without stents) vs.
  any CABG technology (traditional on-pump or
  off-pump bypass)
• Outcomes of Interest The short- and long-term
  objective outcomes (survival, event-free
  survival, non-fatal MIs, CHF, stroke, infections,
  other procedural complications, costs) and
  subjective outcomes (QOL, angina)

16
Synthesizing Survival Curves
DigitizeIt Software
17
Data Synthesis

• Goal To evaluate the comparative effectiveness
  of PCI and CABG at 1-, 6-, 12-, 24-, 36-, and
  60-months post-procedure.
• We computed 2 summary effects for each outcome at
  each time intervals using random effects models
• Measure of absolute risk summary risk
  differences
• Measure of relative risk summary odds ratios

18
Selecting Outcome Metrics
Dead
Alive
PPCI
PCI
1-PPCI
PCABG
1-PCABG
CABG
Absolute Risk Interpretable Stable even when
rates are low Risk Difference PPCI
PCABG Variance (PPCI)(1-PPCI)/NPCI
(PCABG)(1-PCABG)/NCABG
Relative Risk Can be unstable when rates are
low Odds Ratio (PPCI )(1-PPCI)/(PCABG)(1-PCABG)
Variance 1/(PPCI)(1-PPCI)NPCI
1/(PCABG)(1-PCABG)NCABG
19
Observational Studies

• Goals Assess the generalizability of the RCT
  results address questions left unanswered by
  the RCTs.
• Sources Expert advisors literature search for
  RCTs additional, limited MEDLINE and internet
  searches.
• Inclusion criteria
• Large (at least 1000 patients in each arm)
• Provide sufficient information about the patient
  populations and procedures performed to draw
  comparisons with the RCT populations.

20
Todays Talk

• Background on the procedures
• Methodological issues
• Results
• Conclusions and next steps

21
General Description of the Included Studies

• RCTs
• 113 included articles, 22 unique RCTs (See
  handout)
• 9,640 patients 4969 PCI, 4894 CABG
• Most were conducted in Europe/UK
• Only 3 trials performed in the US
• High quality trials except for Seoul
  (?randomized)
• Observational Studies
• 94 included articles, 10 registries
• Generally similar populations to the RCT
  populations (fewer women, more unstable angina
  patients than in the RCTs)

22
RCTs Patient Characteristics
Age Mean 61 years Gender 27
women Race/Ethnicity Overwhelmingly European
ancestry Typical CAD patients who can receive
either intervention (not entire CAD spectrum)
20 diabetes, 50 hypertension,50 ?chol Among
multi-vessel disease (MVD) studies most patients
had 2-vessel rather than 3-vessel disease 40
prior MI, low rates of CHF, good LV
function Very few patients with left main or SVD
23
Results Overall Survival

• Myoprotect and AWESOME enrolled much more acutely
  ill populations than the other 20 trials.

CABG
PCI
Myoprotect
AWESOME
Myoprotect
AWESOME
24
PCI Procedures
Use of stents and adjunctive therapy with
aspirin, clopidogrel/ticlopidine, and heparin was
common in the newer stent, but not in the balloon
trials.
1987-1993 balloon BARI CABRI EAST ERACI GABI La
usanne MASS RITA Toulouse
1998-2003 stents AMIST ARTSAWESOMEERACI
II Groningen Leipzig MASS II

• Myoprotect
• Poland
• Octostent
• Seoul
• SIMA
• SoS

25
CABG Procedures

• LIMA was frequently employed, especially in more
  recent trials.
• Few CABGs were performed off-pump, although a few
  studies used minimally invasive direct coronary
  artery bypass (MIDCAB) incisions and beating
  heart operations in patients with single-vessel
  LAD disease.
• In patients with MVD, surgeons generally bypassed
  more vessels than angioplasty operators dilated.

26
Comparative Effectiveness Results

• Procedural outcomes
• Long-term outcomes
• Subgroup analyses

27
Procedural Survival (Risk Difference)
Heterogeneity Statistics Q-value 20.3, P-value
0.5 I-squared 0.000.
Procedural mortality rare for both procedures no
significant difference between them.
28
Procedural Survival (Odds Ratio)
There is no difference in outcome between
absolute and relative risk metrics.

29
Procedural Survival
1990-1993 (Balloon-era) RD 0.003
1998-2003 (Stent-era) RD 0.005
No difference in procedural survival between
balloon vs. CABG and stent vs. CABG. Also no
comparative difference between SVD and MVD trials.
30
Freedom from Procedural Stroke
Heterogeneity Statistics Q-value 4.9, P-value
0.98 I-squared 0 Odds Ratio Analysis 1.93
(95CI 1.14, 3.29, p0.02)
Procedural strokes (reported by 14 trials) were
significantly higher after CABG than after PCI
(1.2 vs. 0.6). The risk difference 0.6 (95
CI 0.2 to 1, p0.002) and odds ratio 1.9
(95 CI 1.14, 3.29, p0.015) were statistically
homogeneous.
31
Freedom from Procedural MI
Heterogeneity Statistics Q-value 35, P-value
0.009 I-squared 49

• Procedural MIs were slightly higher among CABG
  recipients but not statistically significantly.
  Risk difference 0.1 (95 CI -1.0 to 1.3) and
  odds ratio 1.06 (95 CI 0.72, 1.6). There was
  significant heterogeneity in this outcome.

32
Generalizability of the Procedural Outcomes
Observational Studies

• Procedural risk in RCTs was lower overall than in
  the large registries.
• Society of Thoracic Surgeons operative mortality
  3.2?2.2 (1995?2005) despite patients of higher
  risk profiles
• Hospital Corp. of America (1999-2002) procedural
  mortality was 1.25 for PCI vs 2.63 for CABG and
  that mortality fell for both over the interval
  studied.
• Potential explanation rigorous RCT selection of
  patient, institutions, and operators

33
Comparative Effectiveness Results

• Procedural outcomes
• Long-term outcomes
• Subgroup analyses

34
Long Term Survival

• Overall survival across all randomized trials did
  not differ significantly between CABG and PCI
  between one and five years of follow-up Risk
  difference at each time point was less than 1.
  (See handout.)
• Risk difference between PCI and CABG was smaller
  in the more recent stent trials. Caveat recent
  trials included more patients with single-vessel
  LAD disease and had shorter follow-up.

35
Long-term Survival 11 RCTs
At 1-Year RD -0.001 (95CI -0.009, 0.11) OR
1.1 (95CI 0.86, 1.49)

• At 5-Years
• RD -0.006 (95CI -0.024, 0.11)
• OR 0.9 (95CI 0.78, 1.14)

There is no difference in survival from the 11
RCTs that reported both 1 and 5 year survival
data.
36
Single-vessel vs. Multi-vessel Disease

• 7 trials proximal LAD, 15 trials MVD
• No statistically significant differences in
  comparative survival at any time post-procedure
  between SVD and MVD trials however
• SVD-LAD trials survival was 1 greater for CABG
  recipients (95CI -5 to 2) among balloon
  trials and was 0.1 greater for PCI recipients
  (95CI -4 to 4) among stent trials.
• MVD trials survival 0.6 greater for CABG
  recipients (95CI -1.5 to 0.4) for balloon
  trials but was 1.4 greater for PCI recipients
  (95 CI -1 to 3.8) for the stent trials.

37
Angina Relief
Heterogeneity statistics Q-value 15.0, p-value
0.13, I-squared 33.3.
Angina relief was more complete after CABG than
after PCI at all intervals post-procedure (risk
difference ranges from 5 to 8 odds ratio
ranged from 0.5 to 0.66, plt0.0001 at 1-, 3-, and
5-years).
38
Freedom From Repeat Revascularization

• 1-yr PCI recipients required 24 more repeat
  procedures than CABG recipients (p lt0.0001).
  5-yr This difference climbed to 33
  (plt0.0001).
• Statistically heterogeneous outcome.

39
Freedom from Repeat Revascularization
Balloon Angioplasty
Bare Metal Stents

• At 5-yrs the gap between PCI and CABG narrows
  somewhat among the stent trials compared to the
  balloon trials.

40
Post-procedure MI

• 10 studies reported MIs during follow-up
• Between 1 and 5 yrs post-procedure, the MI rate
  increased among all patients, but at a somewhat
  higher rate for PCI recipients.
• However, the risk differences were small (less
  than 1) (odds ratio ranged from 0.87 to 0.92)
  and did not achieve statistical significance.

41
Quality of Life

• QOL was collected by 11 trials.
• In general, QOL scores were higher among CABG
  patients over 1-3 yrs of follow-up however, many
  trials reported no significant differences
  between PCI and CABG recipients.
• QOL was strongly correlated with angina relief.
  Thus, improvements in QOL were consistent with
  the greater relief of angina by CABG over the
  first few years of follow-up.

42
Costs

• Evaluated in 10 trials.
• Considerable heterogeneity in methods of cost
  determination and health care systems.
• 9 of 10 trials, PCI patients had much lower
  initial costs than CABG patients, but the
  difference in costs between procedures narrowed
  over time.
• In medium to long-term follow-up, PCI-assigned
  patients had only modestly lower costs (about
  5).
• Same trends in both balloon and stent trials.

43
Comparative Cost Data
44
Comparative Effectiveness Results

• Procedural outcomes
• Long-term outcomes
• Subgroup analyses

45
Max
Max is a 79 yo man with well controlled HTN,
hypercholesterolemia, afib, and diet
controlled-DMknowing about the comparative
effectiveness of PCI and CABG by key subgroups
would help his decision making
46
Issues in Subgroup Analyses

• Few studies reported key outcomes for subgroups
  of interest
• Heterogeneity in the outcome metric reported
  (relative risk reduction vs absolute risk
  differences)
• Lower power statistic the common test strategy
  of treatment by interaction p-values

47
Outcomes in Key Subgroups Age

• All RCTs
• -Very few patients 75 yrs were enrolled.
• -There were more procedural complications in the
  older patients, especially stroke.
• Outcomes by age evaluated in BARI, AWESOME, and
  pooled stent trials (ARTS, ERACI-II, MASS-II,
  SoS)
• -BARI patients aged 65 yrs had lower overall
  survival than younger patients. Risk difference
  at 7yrs was larger in the older patients (4.7
  risk difference CABG survival 78.7 vs. PCI
  74.0) than younger patients (2.8 risk
  difference CABG 88.1 vs. PCI 85.3).

48
Outcomes in Key Subgroups Age
Pooled stent trials higher 1-yr incidence of
death, MI and stroke among patients aged 65 yrs
compared with younger patients. Mortality was
lower after CABG in the younger patients, and was
lower after PCI in the older patients, although
there was no statistically significant evidence
of a differential treatment effect by age (age by
treatment covariate interaction test).
49
Outcomes in Key Subgroups Gender

• Gender outcomes reported for BARI, SoS, pooled
  stent trials
• BARI Women had lower overall survival, but only
  at 7 years. The risk difference between PCI and
  CABG was similar in women (3.4 risk difference
  82.6 CABG vs. 79.2 PCI) and in men (3.5 risk
  difference 85.1 CABG vs. 81.6 PCI).
• SoS trial Women had lower QOL at baseline, but
  improved after revascularization. In men, QOL
  improved more with CABG than with PCI, whereas in
  women the improvements were similar with CABG and
  PCI.

50
Outcomes in Key Subgroups Diabetes
Patients with diabetes who have CAD have
substantially higher morbidity and mortality than
patients without diabetes who have CAD. BARI At
5 yrs, CABG recipients with DM had better
survival (80.5) than PCI patients with DM
(65.5). By contrast, no difference in survival
of patients without diabetes between PCI (91.1)
and CABG (91.1). Overall The summary risk
difference at 5 yrs was greater at CABG by only
0.8, but with very wide confidence limits of
-8.3 to 6.6 (odds ratio 0.87 95CI 0.51 to
1.49).
51
1yr Survival for Patients by Diabetes Status
Survival at 1 year for Patients with Diabetes
Survival at 1 year for Patients Without Diabetes
52
5yr Survival for Patients by Diabetes Status
Survival at 5 years for Patients with Diabetes
Survival at 5 years for Patients Without Diabetes
53
Outcomes by Other Risk Factors

• There was no evidence that hypertension, tobacco
  use, renal dysfunction, vascular disease, or
  obesity increased risk differentially in
  PCI-assigned and CABG-assigned patients.

54
Comparative Effectiveness by Extent of CAD

• BARI Better survival for patients with 2-vsl
  rather than 3-vsl disease. Risk difference
  between PCI and CABG at 5-yrs was greater in
  patients with 3-vsl disease (3.9 risk
  difference 84.7 PCI vs. 88.6 CABG) than in
  patients with 2-vsl disease (2.1 risk
  difference 87.6 PCI vs. 89.7 CABG).
• Similar trend reported in the EAST trial and the
  pooled stent-era trials.
• By contrast, RITA-I found very little difference
  in mortality between patients with SVD and MVD
  (mostly 2-vessel) and similar PCI-CABG risk
  differences.
• Overall, these data suggest that mortality in
  patients with 3-vsl disease may be reduced
  somewhat more by CABG than by PCI, and that in
  SVD there is little mortality difference between
  PCI and CABG.

55
Comparative Effectiveness by LV function

• RCTs generally required relatively preserved LV
  function and little CHF. Sick patients excluded.
• Only BARI and AWESOME reported outcomes by LV
  function.
• Worse survival for low LV function patients than
  normal LV function patients. Risk difference
  favored PCI by 0.4 in the patients with low LV
  function (CABG 80.7, PCI 81.1), whereas it
  favored CABG by 4 in patients with normal LV
  function. At 7 yrs, the risk differences favored
  CABG in both the low (1.3) and normal (3.6) LV
  function groups.
• Small and inconsistent differences in the trials
  limit the conclusions about the comparative
  efficacy for PCI and CABG by LV function.
  Ongoing STICH trial is randomizing low EF
  patients to CABG vs. medical therapy (vs. CABG
  with ventricular resection).

56
CABG-specific Factors

• Minimally invasive surgery (MIDCAB)
• small throracotomy incision on a beating heart
• Compared with PCI in 7 small trials
• SVD proximal LAD
• PCI with stents as the comparator
• These trials showed similar outcomes over a
  relatively short follow-up period.
• There are only a few early trials of balloon
  angioplasty and standard CABG (Lausanne, MASS-I)
  in patients with single-vessel disease whose
  results can be compared with these trials.
• In general, the risk differences over one to two
  years of follow-up are comparable.

57
Use of Internal Mammary Grafts

• Use of IMAs ranged from 37 in the early GABI
  study to 90 in the more recent trials.

1 Year
5 Years
Favors PCI
Favors CABG
Favors PCI
Favors CABG
Regression weighted by the sample size We found
that as the of IMA grafts increases, there is a
trend toward the risk difference favoring CABG at
1 yr but not at 5 years (neither association was
statistically significant).
58
Prior Procedures

• Most RCTs excluded patients with prior CABG, but
  1 RCT and several registries compared PCI with
  re-do CABG in patients with a prior CABG.
• AWESOME trial (142 with prior CABG)
• Procedural mortality was higher in the patients
  assigned to re-CABG (8 vs 0), three-year
  mortality was only slightly higher (27 vs 22)
  and not significantly different.
• A similar pattern has been reported by large
  clinical registry studies from Cleveland, Emory,
  and Kansas City procedural mortality was higher
  for re-do CABG than for PCI, but survival at five
  to six years of follow-up did not differ
  significantly.

59
Volume-Outcome Relationships

• Randomized trials have not directly tested for
  differences in clinical outcomes of PCI and CABG
  on the basis of process characteristics.
• There is considerable evidence from observational
  studies that procedural risk of both procedures
  increases significantly in low volume hospitals
  and with low volume operators. This relationship
  remains significant for PCI, even as procedural
  risk has been reduced by the availability of
  coronary stents and adjunctive therapy.

60
Max
Based on the best available evidence, Max chose
CABG. He chose to have it at a high-volume
academic center by a highly experienced surgeon.
His surgery was complicated by a stroke and
prolonged ICU stay. He was discharged to a
skilled nursing facility and died 3 months after
his bypass.
61
Todays Talk

• Background on the procedures
• Methodological issues
• Results
• Conclusions and next steps

62
Conclusions Survival

• Procedural mortality is low for both PCI and CABG
  and does not differ significantly (RD 0.1).
• Long-term mortality was not substantially
  different between PCI and CABG up to 5 yrs after
  the initial procedure (RDs favor CABG over PCI
  between six months and five years by less than
  1).
• 5-yr survival in the balloon-era trials among
  patients with MVD favored CABG, whereas survival
  in the stent-era trials favored PCI, although
  this finding was not statistically significant.

63
Conclusions

• Freedom from angina and repeat revascularization
  strongly favored CABG over PCI.
• The overall survival advantage of CABG over PCI
  among patients with DM was not statistically
  significant, averaging 0.8 in absolute terms at
  5-yrs.
• Only a few trials specifically reported outcomes
  by number of diseased vesselsthese suggest that
  CABG reduces mortality compared with PCI to a
  greater extent in patients with 3-vsl disease
  than in patients with 2-vsl disease.

64
Limitations

• Most RCTs included low prevalence of heart
  failure and LV function was generally well
  preserved, limiting our analyses by LF function.
• Many outcomes, including cognitive function,
  cost, and quality of life outcomes were reported
  by very few studies.
• Most clinical trials have not reported outcomes
  in key subgroups.
• Extended follow-up for the most recent trials was
  not available.

65
Future Research

• Individual patient-level data pooling
• Clinical trials of drug-eluting stents
• Explore minimally invasive approaches to CABG
  with respect to procedure volume and outcomes
• Presently, volume-outcomes studies focus
  exclusively on mortality. Procedural myocardial
  infarction, graft patency, relief of angina, and
  long-term outcomes are also pertinent to decision
  making and should be explored in relation to
  volume levels.

66
Many Thanks

• Collaborators
• Allison Gienger
• Kathy McDonald
• Vandana Sundaram
• Doug Owens
• Marco Perez
• Robin Varghese
• John Kapoor
• Reza Ardehali
• Moira McKinnon
• Christopher Stave
• Ingram Olkin
• Olga Saynina
• Robert Jones
• Funding Agencies AHRQ and VA
• Photo Credit Cynthia Yock