Jeffrey L. Probstfield, MD, for the ACCORD Study Group

1
THE ACCORD TRIAL IS LOWER HBA1C REALLY
DANGEROUS, OR WHY DID PEOPLE DIE IN THIS
STUDY? (and other RECENT STUDIES)

• Jeffrey L. Probstfield, MD, for the ACCORD Study
  Group
• University of Washington Medical Center, Seattle,
  WA

2
(No Transcript)
3
DSMB Recommendation and NHLBI Decision

• NHLBI/NIH decision
• Discontinue intensive glycemia treatment
• Transition all participants to the standard
  glycemia treatment
• No interaction between BP and Lipid Trial
  Components and Glycemia Intervention.
• Continue the BP and Lipid trials
• These trials continue to address important
  questions

(NHLBI Press Release, February 6, 2008)
4
ACCORD Trial Overall Goal

• To determine whether CVD event rates can be
  reduced in people with diabetes by intensively
  targeting three important CVD risk factors
  hyperglycemia, dyslipidemia, and high blood
  pressure.
• Three trials in one research program
• Double 2 by 2 factorial design

5
Glycemia Trial Research Question

• In middle aged/older people with type 2 DM at
  high risk for a CVD event, does a therapeutic
  strategy that targets an A1C lt 6.0 reduce CVD
  event rates more than a strategy that targets an
  A1C between 7.0 7.9 (with the expectation of
  achieving a median level of 7.5)?

6
Glycemia Trial Rationale

• Observational studies supportive
• Each 1 higher A1C associated with 18 greater
  risk of CVD1
• CVD-glucose relationship extends into the normal
  range
• Clinical trials inconclusive2

1. Selvin E, et al. Ann Intern Med.
2004141421-431. 2. Goff DC Jr, et al. Am J
Cardiol. 200799suppl4i-20i.
7
Double 2 X 2 Factorial Design
Lipid
BP
Statin Masked Study Drug
Statin Masked Study Drug
Intensive (SBPlt120)
Standard (SBPlt140)
Intensive Glycemia (A1Clt6)
1178
1193
1374
1383
5128
Standard Glycemia (A1C 7-7.9)
1178
1184
1370
1391
5123
2765
2753
2371
10,251
2362
Primary analyses compare the marginals for main
effects
8
Participant Eligibility

• Stable Type 2 Diabetes for 3 months
• A1C gt7.5 AND lt9 (more meds) OR lt11 (fewer
  meds)
• Age 40-79 previous CVD events OR
• Age 55-79 with
• anatomical ASCVD, albuminuria, LVH OR
• gt 2 CVD risk factors (dyslipidemia, hypertension,
  smoking, obesity)
• BMI lt 45 Cr lt 1.5 (133 uM)
• No frequent/recent serious hypoglycemia
• Able/willing to take insulin, do glucose
  monitoring
• Eligible for BP or Lipid Trial

9
ACCORD Outcomes

• Primary
• First occurrence of nonfatal MI OR Nonfatal
  Stroke OR CV Death
• Secondary/Other
• Each component of 10
• Expanded CVD 10 Revasc HF Hosp
• Total mortality
• Microvascular (nephropathy, neuropathy, eye)
• Eye photo substudy (N 3537)
• HRQL (N 2053) Cost (N 4311)
• MIND cognition, brain volume (MRI)
• Falls/Fractures/BMD

10
Baseline Characteristics
11
Baseline Medications
12
A1C Distribution
Standard Rx Goal
Intensive Rx Goal
13
A1C Distribution 48 Mo.
Standard Rx Goal
Intensive Rx Goal
December 2007
14
ACCORD Glycemia Formulary

• Glargine Insulin
• Aspart Insulin
• 70/30, N, R Insulin
• Exenatide

• Metformin
• Rosiglitazone
• Glimepiride
• Repaglinide
• Acarbose

15
Medications Ever Used During the Trial
of Participants
16
Design of Intensive Glycemia Intervention
Even if the A1C is lt6.0
Rx was reduced in the presence of significant
hypoglycemia.
17
Design of Standard Glycemia Intervention
Decrease only insulin or insulin secretagogues.
18
Changes in Body WeightSince Baseline
Increase
Decrease
19
ACCORD Definition of aSevere Hypoglycemic Episode

• Hypoglycemia requiring medical or paramedical
  attention, AND
• Documented blood glucose lt 50 mg/dl (2.8 mmol/L),
  or
• Prompt recovery with administration of oral CHO,
  IV glucose, or subcutaneous glucagon

Each participants Glucose Diary was reviewed
at each clinic visit to identify the occurrence
of one of these hypoglycemic events
20
Number of Participants With One or More Severe
Hypoglycemia Events Requiring Medical Assistance
(n and )
Cumulative number of events
21
Median A1C and Interquartile Ranges
22
Adverse Events
Motor Vehicle Accident
23
All Cause Mortality
1.41/yr
1.14/yr
HR 1.22 (1.01-1.46) P 0.04
24
Primary Secondary Outcomes
25
Cause of Death
26
Hazard Ratios for Total Mortality by Subgroup
27
Primary Secondary Outcomes
28
Primary Outcome
2.29/yr
2.11/yr
HR 0.90(0.78-1.04) P 0.16
29
Hazard Ratios for Primary Outcome by Subgroup
30
The Question

• Can the observed treatment group difference in
  mortality be explained by the observed
  post-randomization treatment group difference in
  severe hypoglycemia?

31
Severe Hypoglycemia Requiring Medical Assistance
Intensive Group Annual Incidence Rate
3.3 Standard Group Annual Incidence Rate 1.0
32
ACCORD Definition of aSevere Hypoglycemic
Episode

• Hypoglycemia requiring medical or paramedical
  attention
• Documented blood glucose lt 50 mg/dl (2.8 mmol/L)
  or
• Prompt Recovery with administration of oral or IV
  CHO or glucagon

Each participants Glucose Diary was reviewed
at each clinic visit to identify the occurrence
of one of these hypoglycemic events
33
We know
And we know
Intensive Strategy
Higher Mortality
34
So which situation do we have?
Intensive Strategy
Higher Rates of Hypoglycemia
Higher Mortality
And can ACCORDdistinguish these?
No!
35
Background Mortality By Severe Hypoglycemia
Mortality is higher among participants who had
experienced a Severe Hypoglycemic Event
36
Background Mortality By Severe Hypoglycemia
Never Experienced a Hypoglycemic Event
Experienced Hypoglycemic Event
Overall Mortality Rates
1.2 / year
3.3 / year
Intensive Glycemia
1.3 / year
2.8 / year
37
Background Mortality By Severe Hypoglycemia
Never Experienced a Hypoglycemic Event
Experienced Hypoglycemic Event
Overall Mortality Rates
1.2 / year
3.3 / year
Intensive Glycemia
1.3 / year
2.8 / year
Standard Glycemia
1.1 / year
4.9 / year
38
Background Mortality By Severe Hypoglycemia
Never Experienced a Hypoglycemic Event
Experienced Hypoglycemic Event
Overall Mortality Rates
1.2 / year
3.3 / year
Intensive Glycemia
1.3 / year
2.8 / year
Standard Glycemia
1.1 / year
4.9 / year
Again, mortality is higher among participants who
had experienced a Severe Hypoglycemic Event,
regardless of treatment strategy
39
Mortality By Treatment Group
Hazard Ratio (95 CI)
1.22 (1.01, 1.46)
40
Mortality By Treatment Group andSevere
Hypoglycemia
Overall
Never Experienced a Hypoglycemic Event
Intensive Glycemia
1.4 / year (257 Deaths)
1.3 / year (223 Deaths)
Standard Glycemia
1.1 / year (203 Deaths)
1.1 / year (186 Deaths)
Hazard Ratio (95 CI)
1.22 (1.01, 1.46)
1.24 (1.02, 1.50) 1.22 (1.00, 1.48)
Mortality Higher in Intensive Group
Adjusted for baseline factors
41
Mortality By Treatment Group andSevere
Hypoglycemia
Mortality Higher in Intensive Group
Mortality Higher in Standard Group
Interaction P lt 0.01
42
Conclusion I

• Among participants who never had a severe
  hypoglycemic event during follow-up, mortality
  was greater in the intensive group. However,
  among participants who had a hypoglycemic event,
  mortality was greater in the standard group

• Participants who had experienced a severe
  hypoglycemic event were more likely to die

• True for both treatment groups

43
What Is The Relationship Between Prescribed
Medications Mortality In ACCORD?

• Investigating the potential role that medications
  may have played in the mortality imbalance
  continues to be an important consideration for
  ACCORD Investigators.
• This is a difficult task because
• Participants were not randomized to different
  medications
• Choice of medication in any participant was based
  on clinical judgment and characteristics of
  participants

44
Percent of Participants Ever PrescribedGlycemia
Medication
45
Percent of Follow-Up Time WhereMedication Was
Prescribed
46
Relationship Between Participant Characteristics
Prescription of Glycemia Medications

• Examples of how baseline characteristics
  influenced choice of medications.
• Exenatide was less likely to be used if
  participants were already on insulin, as this was
  off label use.
• Insulins were more likely to be used in
  participants with a baseline history of CVD or
  longer duration of diabetes (see next slide)
• Rosiglitazone was less likely to be prescribed in
  those with prior CVD
• In general, these characteristics (longer
  duration, prior CVD) are also associated with a
  greater risk of mortality.

47
Relationship Between Duration of Diabetes and
Post-Randomization Prescription of Medications
15 Years Duration of Diabetes (23 of
Participants)
3.50
3.00
2.50
More Likely
Odds Ratio
2.00
1.50
1.00
0.50
Less Likely
Exenatide
Basal Insulin
Bolus Insulin
Rosiglitazone
Premixed Insulin
48
Relationship Between Prior CVD History and
Post-Randomization Prescription of Medications
Prior CVD (35 of Participants)
3.50
3.00
2.50
More Likely
Odds Ratio
2.00
1.50
1.00
0.50
Less Likely
Exenatide
Basal Insulin
Bolus Insulin
Rosiglitazone
Premixed Insulin
49
Did Combinations Of Medications Lead To The
Increased Mortality In The Intensive Group?

• ACCORD investigators identified several
  combinations that were of interest.
• Estimated the risk of mortality for many
  medications in the presence or absence of a
  second medication. Does the addition of the 2nd
  medication magnify the risk associated with the
  1st medication?
• Because of the high level of interest in the
  combination of rosiglitazone and insulin, we
  present those results on the next slide.

50
Interaction Between Insulin and Rosiglitazone
Mortality HR for Prescription of Rosiglitazone
Stratified By Prescription of Insulin
Basal Insulin
Bolus Insulin
Premixed Insulin
Any Insulin
p0.19
p0.68
p0.93
p0.53
2.25
2.00
1.75
1.50
Hazard Ratio
1.25
1.00
0.75
0.50
0.25
No Insulin
Basal Insulin
Bolus Insulin
Any Insulin Use
No Basal Insulin
Premixed Insulin
No Bolus Insulin
No Premixed Insulin
51
Conclusions -- I

• We have not been able to identify a single agent,
  or combination, that accounts for the imbalance
  in mortality.
• Exenatide ? less mortality, but used rarely and
  more often in Intensive Glycemia group
• Premixed Insulin ? greater mortality, but used
  more often in Standard Glycemia group
• Bolus Insulin ? greater mortality, but no
  difference in mortality hazard ratios by
  randomized group and we dont know if the
  relationship with mortality is a reflection of
  use or the participants to whom it was given
• Approximately a 20 increase in mortality
  associated with Intensive Glycemia even after
  controlling for participant characteristics and
  post-randomization use of glycemia medications.

52
Intensive glycemia strategy discontinued early
after average 3.5 years

• Total mortality (prespecified secondary outcome
  and safety measure)
• 22 higher rate in intensive group
• HR 1.22 (95 CI 1.01-1.46), p 0.04
• Prespecified primary CVD outcome (composite of
  nonfatal MI, nonfatal stroke, fatal CVD)
• 10 lower rate in intensive group
• HR 0.90 (95 CI 0.78-1.04), p 0.16
• DSMB judgment harm outweighs potential benefit
  of intensive treatment, which should be
  discontinued
• NHLBI (sponsor) accepted DSMB recommendation

53
ACCORD Conclusion

• Compared to a strategy targeting A1C levels of
  7-7.9, a therapeutic strategy using currently
  available therapies to target near-normal A1C
  levels in people with longstanding T2DM and
  either CVD or additional CVD risk factors
• Increased mortality
• Did not reduce a composite of major CVD events
  over 3.5 years (primary outcome)
• Mortality results consistent across several
  subgroups
• Suggestion of reduced major CVD events in 2
  subgroups primary prevention and A1C lt8 _at_ BL

54
ADVANCE VADAT STUDY REVIEWS
WHAT TARGETS AND DRUGS FOR DIABETES?
Irl B. Hirsch, M.D. University of Washington,
Seattle, WA
55
Differences in ACCORD/ADVANCE
BASELINE
ACCORD ADVANCE
patients 10,251
11,140 duration DM (yrs) 10 8 Hx
macrovasc. Dz () 35 32 Baseline A1C ()
8.1 7.2 Intervention target A1C
() lt6 lt6.5 insulin Rx ()
77 vs. 55 41 vs. 24 TZD Rx ()
92 vs. 58 17 vs. 11 Outcome
(intensive vs. standard) Median A1C _at_ study end
6.4 vs. 7.5 6.4 vs. 7.0 DEATH
any cause 5.0 vs. 4.0 8.9 vs. 9.6
Plt0.05
NEJM 2008358, 2630
56
ADVANCE
57
ADVANCE Primary Outcomes
58
ADVANCE Primary Outcomes
59
ADVANCE Secondary Endpoints

• All-cause mortality P NS
• Total renal events 11 RR with intensive, P lt
  0.001
• Eye events P NS
• CHF, PVD, neuropathy P NS

60
VA DAT

• Participants 40 w prior CVD 80 w HTN, vast
  majority obese average age 60
• Participants had to have failed simple therapy
• Treatment intensive group most on 2-3 orals plus
  insulin (90 on insulin)
• Achieved goals for lipids (LDL 78 mg/dL), BP
  (127/70) these were background Tx
• BL A1C mean 9.5
• Achieved A1Cs standard 8.4, intensive 6.9 -
  delta 1.5

UNPUBLISHED
61
VA DAT

• Reduction in CV events composite outcome,
  events and procedures not statistically
  significant
• All components of the CVD composite outcome
  favor good glucose control, except CV death,
  where there was a .slight insignificant
  increase
• CVD events standard 263, intensive 231 rates
  lower than predicted
• Microvascular outcomes not reported analyses
  pending, to be reported in September in Rome
• Telling quote We thinkA1C is a stronger
  marker for microvascular complicationsthan for
  macrovascular complications

UNPUBLISHED
62
Candidate Mechanisms TGC and CVD Events

• Hypoxia (remember the PDR story!)
• Hypoglycemia (arrhythmias, brain dysfunction,
  vasoconstriction, new data leading to DAN)
• Obesity (3 drugs resulting in weight gain)
• Glucose variability in long-standing diabetes
  (insulin deficiency)

63
Big Picture Messages

• T1 and T2DM early meticulous glucose control can
  prevent microvascular and neuropathic
  complications
• T1DM early meticulous glucose control appears to
  prevent CVD many years later
• T2DM early meticulous glucose control does
  appear to prevent both micro- and macrovascular
  disease in T2DM

64
The Benefit of Early Aggressive Glycemic Control

• Metabolic memory
• Legacy effect

65
Big Picture Message

• T2DM patients with known CVD or long durations
  of DM may be harmed by meticulous control
  although the mechanism(s) for this are not known,
  the leading candidate mechanism is hypoglycemia

66
More Big Picture Messages

• T1DM impact of glycemia on microvascular disease
  not present after 20-25 years (probably true for
  T2DM too)
• After long duration of either T1 or T2DM (or
  known CVD), it appears BP, LDL-C and ASA use
  better predict CVD mortality than A1C
• Impact of hypoglycemia is not consistent between
  populations (under 5 year-olds, geriatrics,
  inpatient)

67
SO WHAT A1C TARGETS?
68
My Take, At Least While We Are Awaiting ADA/AACE
Consensus Statements on T2DM Targets

• lt 10 years T2DM AND no CVD Target at least lt 7
• 1st line metformin
• 2nd line SFU, sitagliptin, exenatide, basal
  insulin (A1C lt 9)
• 3rd line physiologic insulin therapy
• 10-15 years T2DM AND no CVD
• No change from above but this population will be
  more likely to require insulin to reach A1C
  target

69
Possible Strategy

• gt 15 years T2DM OR known CVD 7-7.5 A1C
• Drugs with less risk of hypoglycemia
• Metformin, SFU unlikely to be effective with
  longer durations of DM
• Little data for TZDs, exenatide, sitagliptin
• Greatest risk of hypoglycemia with insulin, but
  also greatest likelihood of efficacy to
  consistent A1C levels
• Less hypoglycemia with basal insulin alone, but
  some prandial insulin required as duration of DM
  and A1C increases
• Dont use basal insulin to replace prandial needs!

70
Conclusions

• The recent studies do not negate the years of
  research from other clinical trials
• Different populations appear to have different
  A1C targets
• It appears the same in the inpatient population!
• It is difficult to recommend a generalization of
  one drug vs. another (depending on the situation)
  as there are so many variables and little
  clinical trial data to guide us
• General hypoglycemia, weight gain, pregnancy,
  cost
• Specific GI tolerability, edema, bone fx,
  increase CVD risk (?)

71
Conclusion

• Insulin is always an option, is under-utilized,
  and needs to be used in a physiologic manner in
  patients with severe insulin deficiency
• In patients with known vascular disease, even
  more modest A1C targets require the use of
  insulin analogues (as opposed to human insulins)
  due to the consistent data showing less
  hypoglycemia even though there are no differences
  in A1C.