In Defense of Avandia

(rosiglitazone)

Hal Roseman, MD, MPH, FACC, FACP

1
Hal M. Roseman, MD, MPH, FACC, FACP
  B.A. (Social and Political Theory)- University of
Pennsylvania

  M.D.- University of Tennessee

  M.P.H. (Epidemiology of Cardiovascular Disease)- Yale

University

  Cardiology Training:

  Brown University Affiliated Hospitals

  Massachusetts General Hospital

  Specialist, American Society of Hypertension

  Fellow, National Lipid Association

Hal M. Roseman, MD, MPH, FACC, FACP
  Conflict Declaration:

  Speaker Bureaus: GSK, Abbott, Novartis, Merck

  No stock ownership

  No advisory positions

  This presentation represents my own opinions. I did not

receive financial support from GSK to make this
presentation.
http://www.time.com/time/health/article/0,8599,1623580,00.html?xid=feed-cnn-topics
No defining evidence to support a causal link between
rosiglitazone and adverse cardiovascular events.

AHA/ACC Science Advisory1 (2010):

“In summary, an association
between rosiglitazone and IHD
outcomes has not yet been firmly
established.”

FDA2 (2007):
“The available data on the risk
of myocardial ischemia are
inconclusive.”
1.  Kaul S, et. al., Circulation. 2010;121:1868-1877
2.  FDA Issues Safety Alert on Avandia. www.fda.gov/bbs/topics/NEWS/2007/NEW01636.html
2009 AACE/ACE Glycemic Control Algorithm
  American Association of Clinical Endocrinologists (AACE) and the American
College of Endocrinology (ACE) released a new algorithm for the treatment of
patients with type 2 diabetes in October 2009

  AACE/ACE consensus panel has taken into consideration the results of several,
large clinical trials, including RECORD, and continues to recommend the use of
AVANDIA® (rosiglitazone maleate)

  The authors state “Although some studies have been controversial, recent
clinical trials ─ADVANCE , VADT, and ACCORD─showed no increased risk of
mortality associated with rosiglitazone…”

  The authors also discuss the known side-effect profile of the thiazolidinedione
(TZD) class, including the risk of edema and congestive heart failure

AVANDIA continues to be included in the AACE/ACE algorithm

for appropriate patients with type 2 diabetes

Rodbard HW, et al . Endocrine Practice. 2009;15(6):540-559.

6
Top Five Reasons for Keeping
Avandia on Open US Formulary.
5. Given the benefits of
durable glucose control in
preventing microvascular
disease, Avandia’s
demonstrated ability to
control blood sugar for five
years is unique among all
other diabetic medications.
 ADOPT & RECORD:
Similar Patterns of Durable A1c Control1-3
8.0 RECORD*†
8.0 ADOPT 7.8 Background SU
Metformin

7.8 7.6
7.6 Sulfonylurea 7.4 AVANDIA
7.4 7.2
Metformin 0 1 2 3 4 5
7.2 Time (years)
8.0
Sulfonylurea
7.0 Background MET
7.8
6.8 AVANDIA
7.6
6.6
7.4
6.4 AVANDIA
7.2
6.2
0 1 2 3 4 5 7.0
Time (years)
6.8
0 1 2 3 4 5
*This study was not designed to show superiority in glycemic control. Time (years)
†Model-adjusted mean A1c by background strata.

1. Kahn SE et al. N Engl J Med. 2006;355:2427–2443. 2. Data on File, GlaxoSmithKline. 3. Home PD, et al. Lancet.
2009;373:2125–2135. 9
 Longest Glycemic Control Monotherapy Trial
with Pioglitazone in Patients with T2DM*

12
11.5
11
FPG (mmol/L)

10.5
10
9.5 Gliclazide
9
Actos
8.5
8
0 8 24 52 78 104
Weeks of Treatment
Note: No studies are available evaluating glycemic control with pioglitazone for greater than two years.
FPG = fasting plasma glucose; T2DM = type 2 diabetes mellitus.
*P <0.01 for all timepoints between week 52 and week 104.
Adapted from Tan et al. Diabetes Care. 2005;28:544–550. 10
4. In terms of validity
of data, randomized
clinical trials are the
gold standard and
trumps that of
metanalyses (e.g.,
Steve Nissen) or
observational data
sets (e.g. Graham)
 “There is simply no serious scientific
alternative to the generation of large-
scale randomized evidence.”
– R Peto
Professor of Medical Statistics and Epidemiology
University of Oxford
Oxford, United Kingdom
Contributed to the development
of the meta-analysis method

Peto R, Collins R, Gray R. J Clin Epidemiol 1995; 48: 23-40

Peto R et al. J Clin Epidemiol. 1995;48:23–40.

 Independent Diabetes Outcome Trials:
Findings with Regard to AVANDIA
  The investigators of VADT and ACCORD, 2 independent,
long-term CV outcome studies involving more than 20,000
years of patient experience with rosiglitazone in a high-risk
population of patients with type 2 diabetes, reported that
AVANDIA was not related to an increased risk of mortality1,2

  Although VADT and ACCORD were not prospectively

designed to assess the safety of AVANDIA, the mortality
results for AVANDIA are consistent with results from other
long-term studies (DREAM, ADOPT, RECORD) with
AVANDIA1-4

1. American Diabetes Association. http://www.diabetes.org/for-media/pr-intense-blood-glucose-control-

yields-no-significant-effect-on-cvd-reduction.jsp. Accessed April 1, 2010.
2. ACCORD Study Group. N Engl J Med. 2008;358:2545–2559.
3. Prescribing Information for AVANDIA.
4. Duckworth W et al. N Engl J Med. 2009;360(2):129–139 13
ACCORD: Avandia not linked to increase deaths
seen in the aggressively treated arm
No Significant Difference in Myocardial Infarction* in
GSK Long-term Prospective Studies
No. of patients
Study Treatment HR (95% CI)
with event
ADOPT‡
RSG (N = 1456) 20 (1.4%)
vs MET (N = 1454) 17 (1.2%) 1.21 (0.64, 2.32)
vs SU (N = 1441) 15 (1.0%) 1.20 (0.62, 2.35)
DREAM†
RSG (N = 1325) 5 (0.4%)
7 (0.5%) 0.71 (0.23, 2.24)
vs Placebo (N = 1321)

RSG + Ramipril (N = 1310) 12 (0.9%)

vs Ramipril (N = 1313) 5 (0.4%) 2.41 (0.85, 6.84)

0.5 1.0 5.0

Favors RSG Favors control

*Adjudicated fatal and nonfatal myocardial infarction plus sudden death;

†Events were adjudicated during the study; ‡Post-study adjudication.

Prescribing Information for AVANDIA.

Data on file, GlaxoSmithKline.
No Significant Difference in MACE (CV Death, MI, Stroke)
in GSK Long-term Prospective Studies
No. of patients
Study Treatment
with event HR (95% CI)
ADOPT†
RSG (N = 1456) 35 (2.4%)
vs MET (N = 1454) 36 (2.5%)
1.00 (0.63, 1.59)
vs SU (N = 1441) 28 (1.9%)
1.11 (0.67, 1.82)
DREAM*
RSG (N = 1325) 15 (1.1%)
vs Placebo (N = 1321) 14 (1.1%) 1.07 (0.51, 2.21)

RSG + Ramipril (N = 1310) 18 (1.4%)

vs Ramipril (N = 1313) 9 (0.7%) 2.01 (0.90, 4.48)

0.5 1.0 5.0

Favors RSG Favors control

*Events were adjudicated during the study.

†Post-study adjudication for CV death and myocardial infarction; not adjudicated for stroke.

Prescribing Information for AVANDIA.

Data on file, GlaxoSmithKline.
 No Significant Difference in Total Mortality
in GSK Long-term Prospective Studies
No. of patients
Study Treatment
with event HR (95% CI)
ADOPT
RSG (N = 1456) 12 (0.8%)
vs MET (N = 1454) 15 (1.0%) 0.82 (0.39, 1.76)
vs SU (N = 1441) 21 (1.5%) 0.51 (0.25, 1.04)
DREAM
RSG (N = 1325) 15 (1.1%)
vs Placebo (N = 1321) 17 (1.3%) 0.87 (0.44, 1.75)

RSG + Ramipril (N = 1310) 15 (1.1%)

vs Ramipril (N = 1313) 16 (1.2%) 0.94 (0.47, 1.90)

0.2 0.5 1.0 5.0

Favors RSG Favors control

Prescribing Information for AVANDIA.

Data on file, GlaxoSmithKline.
 VADT: Rosiglitazone Use is not
associated with Increased CV Risk
VADT: Effect of Rosiglitazone dosage
on Time to First Myocardial Infarction

Moritz T. Presentation at the 68th Scientific Sessions of the ADA: June 8, 2009 (San Francisco, CA)
 VADT: Effect of Rosiglitazone dosage
on Time to MACE

•  Baseline covariates: age, prior event, diabetes duration, baseline HbA1C, minority, SBP, total cholesterol, HDL
** baseline and time dependent covariates: age, prior event, # of Insulin, diabetes duration, minority, total cholesterol, HDL, statins
 BARI-2D: Rosiglitazone Effect on
Cardiovascular Events
“Our observations from BARI 2D do not
suggest a significant cardiovascular hazard
and may suggest a potentially beneficial effect
on ischemic cardiovascular events associated
with treatment with rosiglitazone among
patients with type 2 diabetes and established
coronary artery disease like those treated in
the trial"

Dr Bach, Presented at 70th Scientific Sessions of the

ADA ; June 2010 (Orlando, FL.)
 BARI-2D SubAnalysis:
Rosiglitazone vs. Non-TZD Treatment
BARI-2D Trial: CV Events during Treatment
with Rosiglitazone Plus 3 Months vs. No TZD

Unadjusted Outcomes, Mean Follow-up 4.5 years

 BARI-2D:
Metabolic Benefits of Insulin Sensitization Treatment

In our study, plasma insulin levels were

consistently lower over time in patients in the
insulin sensitization group…Moreover, the
insulin-sensitization strategy was associated with
fewer severe hypoglycemic episodes, less
weight gain, and higher HDL levels than those in
the insulin-provision strategy. These data may
suggest that insulin sensitization is preferable for
patients with type 2 diabetes and coronary
disease.
N Engl J Med 2009;360:2503-15.
3. Meta-analysis, by
their nature, is prone
to methodological
problems, as
exemplified by the
original Nissen
meta-analysis.
Article published online on May 21, 2007
N Engl J Med 2007;356
Meta-analytic Odds Ratios for Myocardial
Infarction and Cardiovascular Death

Diamond GA, Bax L, Saul K. Ann Intern Med. 2007;147:578-581

Uncertainty of Peto Method: Fixed vs
Random Effects
2. There has not been
put forth a convincing
hypothesis why Avandia
would cause cardio-
vascular harm, an
important omission.
 ADOPT: A Diabetes Outcome Progression Trial
Long-term Lipid Data from ADOPT
•  Percentage of patients on lipid-lowering therapy at baseline (including statins) –
26.0% AVANDIA, 25.9% MET, 25.7% SU1
•  Statin use at 4 years – 51.5% AVANDIA, 43.5% MET, 40.2% SU1
•  Lab values assessed at 4 years (48 months)1

Baseline Lipids: LDL: 119-121 (98-144);

HDL: 46.5-47.3 (39.0-55); Trigs: 156-163 (112-233)
Results shown are adjusted geometric mean change from baseline.
*Percentage change calculated based on log transformed data.
1. Kahn SE et al. N Engl J Med. 2006;355:2427–2443. 2. Data on File, GlaxoSmithKline.
 RECORD: Rosiglitazone Evaluated for Cardiac
Outcomes and Regulation of glycemia in Diabetes
Lipid Data in Patients Taking AVANDIA
  Baseline statin use – 18.0% RSG vs 19.2% active control
  Statin use at 5 years – 55.2% RSG vs 46.0% active control
  Lab values assessed at 5 years

TG 1.69 mmol/L = 150 mg/dL; HDL 1.03 mmol/L = 40 mg/dL; LDL 2.59 mmol/L = 100 mg/dL.

Results shown are mean results for lipid parameters.

*Estimates of 5-year changes obtained with a baseline-adjusted repeated-measures model for all patient data (and P values
for treatment difference).
†Lipids were not measured after initiation of any insulin therapy.

Home PD et al. Lancet. 2009;373:2125–2135. 32

PROactive: PROspective pioglitAzone
Clinical Trial In macroVascular Events
Lipid Data in Patients Taking Pioglitazone
•  Baseline statin use – 43% PIO vs 43% placebo
•  Statin use at study end – 55% PIO vs 56% placebo
•  Lab values assessed at final visit, ≈34.5 months (2.9 years)

Baseline % Change
PIO placebo PIO placebo P

TG (mmol/L) 1.8 1.8 -11.4% +1.8% <0.0001

HDL (mmol/L) 1.1 1.1 +19.0% +10.1% <0.0001

LDL (mmol/L) 2.9 2.9 +7.2% +4.9% 0.003

TG 1.69 mmol/L = 150 mg/dL; HDL 1.03 mmol/L = 40 mg/dL; LDL 2.59 mmol/L = 100 mg/dL.

Results shown are median results for lipid parameters.

TG = triglycerides; HDL = high-density lipoprotein; LDL = low-density lipoprotein.
Dormandy JA et al. Lancet. 2005;366:1279–1289.
Both Avandia and Actos have demonstrated
significant effects on known cardiovascular surrogate
markers

•  Carotid intima thickness regression

•  Albumin reduction

•  BP reduction
PPARγ activation: Consistent reduction
in carotid atherosclerosis
Patients (n)
Study (year) Treatments duration Δ IMT (mm)
Minamikawa Troglitazone 400 mg Type 2 diabetes ↓0.080, troglitazone
(1998) Usual care (n = 135) ↑0.027, usual care
6 mos P < 0.001

Koshiyama Pioglitazone 30 mg Type 2 diabetes ↓0.084, troglitazone

(2001) Usual care (n = 106) ↑0.022, usual care
6 mos P < 0.001

Sidhu Rosiglitazone 8 mg Stable CAD ↓0.012, rosiglitazone

(2004) Placebo (n = 92) ↑0.0031, placebo
12 mos P = 0.03
Langenfeld Pioglitazone 45 mg Type 2 diabetes ↓0.054, pioglitazone
(2005) Glimepiride 2.7 mg (n = 173) ↓0.011, glimepiride
6 mos P < 0.001

Minamikawa J et al. J Clin Endocrinol Metab. 1998;83:1818-20.

Koshiyama H et al. J Clin Endocrinol Metab. 2001;86;3452-6.
Sidhu JS et al. Arterioscler Thromb Vasc Biol. 2004;24:930-4.
Langenfeld MR et al. Circulation. 2005;111:2525-31.
Effects Of Rosiglitazone On CV Biomarkers
In Diabetics And Non-Diabetics
Rosiglitazone

PPARγ

↓ sdLDL ↓ Diastolic BP ↓ CRP, IL-6 (?)*

↓ Insulin Resistance
↑ HDL2 ↑ Vascular Reactivity* ↓ sCD40L*
↓ Glucose/HbA1c
↔/↓ TC/HDL3 ↓ Microalbuminuria* ↓ *ROS, MCP-1, P47phox*
↓ Insulin & Proinsulin
↓ MMP-9* ↓ TNFα*
↓ E-selectin*
↓ PAI-1* ↓ ADMA*
↓ NFκB*

↑ NO In Human Skin*
↑ Adiponectin*

Improved Improved Improved Reduced

Fibrinolytic Activity* Lipid Profile Vascular Function* Inflammation*

? CV and DM Outcomes, Atherosclerosis ?

* Freed et al. Diabetologia. 2000;43(suppl 1):1267. Data on file, GlaxoSmithKline. Mohanty et al. Diabetes. 2001;
50(suppl 2):A68 (Abstract 276-OR). Stuhlinger et al. JAMA. 2002;287:1420-1426. Brunzell et al. Diabetes. 2001;
50(suppl 2):A141 (Abstract 567-8 and poster). Chu et al. Diabetes. 2002;51(suppl 1):(Abstract 553-P). Agrawal et al.
Diabetes. 2002;51(suppl 1):A92 (Abstract 372P). Vinik et al. Diabetes. 2002;51(suppl 1):A82 (Abstract 333P).
 LUMEN
Endothelial Injury LDL
Adhesion Molecules
PAI-1 Cytokines

MCP-1 Colony
Stimulating
Factors
Oxidization
Monocyte of LDL INTIMA
Cholesterol Efflux !!
Transformation

Macrophage Foam Cell

Growth
Factor: Mitotic Angiogenic
Migration Factors: MMP MEDIA
PDGF

Effect of Thiazolidinediones
ASMC on Atherosclerosis
1. The alternative
therapy, Actos
(pioglitazone), has
insufficient data to be
trustworthy as the only
PPARγ agonist on the
US formulary.
 PROactive: the only Long-Term Study with
Cardiovascular Safety Endpoints: Design and Results
Design Results
  5238 patients   Failed to achieved primary
  100% previous cardiovascular endpoint
disease
  Principal secondary endpoint
  Average follow-up 2.9 years; (MACE*) 301 events PIO vs
15,060 patient-years 358 events placebo
  >30% insulin-treated at baseline   417 CHF events PIO vs 302
  Pioglitazone + treatment vs placebo events placebo (non-
+ treatment design – double blind adjudicated)

  CHF not included in primary and   Paper did not report fatal and
secondary endpoints non-fatal MI together

*For PROactive, MACE = all-cause death, non-fatal MI (excluding silent MI), and stroke.
Dormandy JA et al. Lancet. 2005;366:1279–1289.
 According to the AHA and ACC, meta-analyses and
observational studies are not sufficient to evaluate clinical
differences between rosiglitazone and pioglitazone

Substantial differences between the pioglitazone and rosiglitazone meta-

analyses exist, eg, placebo-controlled versus active-controlled trials,
patient demographics, and treatment duration. Each of these factors
potentially can have a material impact on outcomes. This type of indirect
comparison is potentially misleading, may result in conflicting results
depending on the end points compared, and generally should be
avoided. Healthcare databases used in observational studies are limited
by bias and confounding, and therefore, they are not particularly well
suited for drawing definitive conclusions to impact policy or clinical
practice recommendations.

Kaul S, et. al., Circulation. 2010;121:1868-1877.

 AHA/ACC Science Advisory: Thiazolidinedione
Drugs and Cardiovascular Risks

However, there remains an inadequate foundation

of randomized clinical trials to properly judge the
safety or efficacy of either agent with respect to
IHD events.

Kaul S, et. al., Circulation. 2010;121:1868-1877.

END