Setting the Record Straight

1
Setting the Record Straight
2
Major ALLHAT Findings

• CHD risk not improved for any of the 3 newer
  agents compared with chlorthalidone
• Total mortality was similar for the 4 groups
• Diuretic superior in preventing one or more major
  forms of CVD, including stroke and heart failure
• Subgroups consistent except stroke, combined CVD
• Heterogeneity in L / C comparison by ethnicity
  greater reductions in Blacks
• Diuretics drug of choice for initial therapy of
  HTN and should be included in multidrug regimens

3
Setting the Record Straight Study Design

• How could ALLHAT test first-step therapy, given
  the studys inclusion criteria and lack of a
  washout period?

4
Testing First-Step Therapy The Ideal Trial

• Include all hypertensive patients
• Low and high risk
• Treated (with washout) and untreated
• BUT
• Require more patients
• More complex
• Unaffordable

5
Testing First-Step Therapy ALLHAT

• Practice-based trial mirrors community treatment
  of hypertension
• Obtained sufficient patients
• Captures diversity of patients
• High risk patients assure adequate numbers of
  outcome events
• No washout, except for ß-blockers

6
Setting the Record Straight Study Design

• Why were diuretics and calcium-channel blockers
  avoided as second-step drugs?

7
Second-Line Drugs

• Second- and third-line drugs available for BP
  control
• Discouraged step-up from same class as any of the
  first-step agents unless compelling indications
• Odd that ß-blocker a step-up agent for ACEI?
• Reserpine, clonidine, hydralazine also provided
  as step-up therapy in addition to ß-blocker
  different mechanisms of action than first-step

8
Second-Line Drugs BP Control

• BP control with ALLHAT regimen more than twice
  that at entry
• Exceeded that observed in 3rd NHANES

9
Setting the Record Straight Study Conduct

• Doesnt the attrition rate necessarily bias the
  conclusions?

10
Study Conduct Attrition

• Mean length of follow-up 4.9 years
• 99 of expected person-years were observed
• 97.1 of participants had known vital status
  during closeout period
• Sensitivity analyses consistent with trials
  published conclusions

11
Setting the Record Straight Study Conduct

• Wasnt the outcome ascertainment process flawed
  since end points were not systematically reviewed
  by a panel of experts?
• Arent the secondary outcomes soft end points?

12
Study Conduct Endpoint Ascertainment

• Not feasible to systematically verify all
  endpoints
• 11,000 CVD end points during follow-up
• AHT double-blind ? no bias for or against any
  treatment when reporting and classifying
  endpoints
• LLT not double-blind ? potential bias for all
  nonfatal outcomes ? secondary endpoints for LLT
  soft data

13
Study Conduct Endpoint Ascertainment

• Investigators trained per definitions detailed in
  Manual of Operations
• Review of all end points at ALLHAT Clinical
  Trials Center by medical reviewers.
• Verified investigator-assigned diagnoses using
  death certificates discharge summaries

14
Study Conduct Endpoint Ascertainment

• Random 10 subset of CHD stroke more detailed
  information collected reviewed by Endpoint
  Subcommittee
• 90 agreement for primary outcome (CHD)
• 84 agreement for stroke
• Smaller one-time sample of HF cases
• 85 agreement
• Rates of agreement similar across treatment
  groups.

15
Setting the Record Straight Conclusions and
Interpretations

• Why do the authors emphasize the secondary
  outcome results?

16
Conclusions Interpretations Primary vs
Secondary Outcomes

• Identification of primary outcome assures
  statistical power to test question related to
  that end point
• Primary outcome essentially identical in all
  treatment groups.
• Other important predefined clinical outcomes
• Public health viewpoint, all major clinical
  outcomes are worth examining
• E.g., Total mortality

17
Setting the Record Straight Conclusions and
Interpretations

• Are the heart failure findings real?
• Cant all or most of the heart failure findings
  be explained by the use of antihypertensive
  medications, such as diuretics and CCBs, before
  entry into ALLHAT?

18
Conclusions Interpretations Heart Failure
Validity

• First validity sample - 85 agreement in 39 cases
• All HF hospitalizations and deaths 3031 cases
  in 2091 patients
• All relevant materials collected, 2 reviewers per
  case (blinded to treatment group)
• ALLHAT and Framingham criteria, reviewers
  judgment
• Confirmed 70-84 of cases in each treatment
  group, depending on criteria used
• Analysis using only confirmed cases confirmed
  original ALLHAT findings regarding HF

19
Conclusions Interpretations Early Divergence
of HF Differences

• Divergence continued after 1 year for doxazosin
  amlodipine vs chlorthalidone
• For lisinopril vs chlorthalidone, curves
  converged between 6-7 years

20
Conclusions Interpretations Suggested Reasons
for Divergenceof HF Curves

• Precipitation of edema with amlodipine?
• Unmasking of edema upon withdrawal of diuretics
  at entry?
• Central review algorithm for HF disallowing
  peripheral edema
• Did not alter HF confirmation rate
• Did not alter treatment group differences

21
Conclusions Interpretations HF Findings vs
Meds at Entry

• IMS data 1994-1998 (ALLHAT recruitment)
• U.S. hypertensives taking diuretics decreased
  from 30 to just over 20
• Central review of HF cases
• No interaction of study treatment with pre-entry
  diuretic use

22
Conclusions Interpretations HF vs 2nd and 3rd
line drugs

• Addition of 2nd and 3rd line drugs probably
  contributed to lessening of the divergence 6-12
  months after randomization
• Open-label diuretics, ß-blockers, ACEI
• Excess risk with doxazosin as monotherapy reduced
  but not eliminated after 1 year
• Greatest differential in participants with
  controlled BP difference not explained by BP
  differential

23
Conclusions Interpretations HF vs Total
Mortality

• ? HF ? ? total mortality?
• 9 excess cases of fatal HF for lisinopril
• lt1 of all deaths
• 39 fatal HF for amlodipine, 3 of deaths
• Differences unlikely to be detected

24
Setting the Record Straight Conclusions and
Interpretations

• Cant all or most of the outcome findings
  (especially the differential ethnicity subgroup
  findings for stoke) be explained by the observed
  blood pressure differences among the treatment
  groups?

25
Conclusions Interpretations Blood Pressure
Differences

• Goal achieve equivalent BP control in all 4
  groups
• Mean decrease in BP not a declared outcome
• Chlorthalidone-based regimen the most effective
  in reducing clinical outcomes and, to a small
  degree, in lowering BP

26
Conclusions Interpretations Blood Pressure
Differences
If a given agent is less effective in reducing
clinical events unless it is combined with
another agent like chlorthalidone to lower BP,
not clear why treatment would be started with
anything other than diuretic
27
Conclusions Interpretations Blood Pressure
Differences

• ? achieved SBP ? ? in CV findings?
• Meta-regressions of BP differences on trial
  results
• True to some extent, except for HF

28
Conclusions Interpretations Blood Pressure
Differences

• ? BP for amlodipine vs chlorthalidone, and for
  lisinopril vs chlorthalidone in non-Black
  participants ? 1 mm Hg
• Expect no / negligible effect on CV events
• HF higher with amlodipine (38) and with
  lisinopril (15) than with chlorthalidone
• Larger differences in Black participants
• 4 mm Hg SBP in lisinopril vs chlorthalidone
• Explains lt ½ of observed higher risk for stroke
  (40) and HF (32)

29
Setting the Record Straight Conclusions and
Interpretations

• Doesnt the increased incidence of new diabetes
  in the chlorthalidone group portend greater
  long-term cardiovascular risk for patients taking
  this drug?

30
Conclusions Interpretations Incident Diabetes

• Incident diabetes not a pre-specified outcome
• Thiazide diuretics ? small increase in serum
  glucose (3-4 mg/dL) in short term
• Consistent with other literatuve
• Results for major outcomes consistent by baseline
  diabetes status

31
Conclusions Interpretations Incident Diabetes

• ? in serum glucose did not lead to ? CV events or
  ? total mortality during the trial
• Patients in doxazosin group had ? mean glucose
  compared to chlorthalidone
• Did not translate in better CV reduction for
  doxazosin

32
Conclusions Interpretations Incident Diabetes

• Thiazide-induced diabetes can probably be
  prevented or reversed
• Maintenance of potassium balance
• Adequate weight control
• Increased physical activity
• Caution when using ß-blockers in combination
  therapy

33
Conclusions Interpretations Incident Diabetes

• Long follow-up for ALLHAT, avg. 4.9 years
• Cannot predict outcomes beyond trials duration
• Applies to any clinical trial
• Lack of evidence that a result will hold up
  decades after trial ends does not prove that a
  different outcome will result
• Does thiazide-induced diabetes carry same
  prognosis as naturally-occurring diabetes?

34
Setting the Record Straight Conclusions and
Interpretations

• Diuretics themselves may be cheaper, but does the
  cost of management with diuretics translate into
  less expensive therapy?

35
Conclusions Interpretations Cost of
Antihypertensive Management

• Cost subordinate to safety efficacy
• Still should be considered in selection of
  antihypertensive agents
• Could have major impact on health care
  expenditures in U.S.
• Diuretic use declined from 56 of prescriptions
  in 1982 to 27 of prescriptions in 1992
• 3.1 billion in savings on drugs costs if
  diuretic use had remained at 1982 levels

36
Conclusions Interpretations Cost of
Antihypertensive Management

• Cost effectiveness analyses for ALLHAT are
  underway
• Preliminary analyses suggest costs driven by drug
  acquisition
• Cost for monitoring K and glucose not proven to
  be more than that required during treatment with
  ACEI or in routine care of patients with risk
  factors.

37
Setting the Record Straight Conclusions and
Interpretations

• Can the findings be extrapolated to drugs within
  class?

38
Conclusions Interpretations Extrapolation to
Drug Classes

• For a-blockers, ACE inhibitors, dihydropyridine
  CCBs, extrapolation seems reasonable
• Chlorthalidone ? thiazide diuretics ? HCTZ?
• MRFIT mortality trends less favorable at clinics
  where HCTZ favored over chlorthalidone
• Based on post hoc subgroup analysis
• Based on group identifier (clinic) rather than
  patients results did not hold up at patient
  level

39
Conclusions Interpretations Extrapolation to
Drug Classes

• Data from other studies (except MRFIT) using
  various thiazide-type diuretics suggest similar
  benefit in CVD prevention
• Chlorthalidone
• HCTZ
• Indapamide
• Bendrofluazide

40
Setting the Record Straight Conclusions and
Interpretations

• Why do the findings from ALLHAT and the Second
  Australian National Blood Pressure Study
  seemingly conflict?

41
Conclusions Interpretations ALLHAT vs ANBP2

• Second Australian National Blood Pressure Study
• Practice-based open-label trial
• Diuretic-based vs ACEI-based treatment
• Recommended HCTZ, enalapril
• 6083 participants, 65-84 years of age
• Followed for a mean of 4.1 years

42
Conclusions Interpretations ALLHAT vs ANBP2

• Primary endpoint - composite of all CV events
  (initial recurrent) plus all-cause mortality
• Results marginally favored ACEI
• RR 0.89 (0.79 1.00, p0.05)
• First CV event or death, p0.06
• First CV event, p0.07

43
Conclusions Interpretations ALLHAT vs ANBP2

• Frohlich NEJM. 20035192-5 - samples studied,
  specific drugs used
• 2X CV events in ALLHAT as participants in ANBP2
• ALLHAT double-blind vs ANBP2 PROBE design
• increased potential for bias in ANBP2
• Results consistent if upper confidence limit for
  relative risks in ANBP2 compared with estimates
  in ALLHAT

44
Limitations Expectations

• New drugs have been or will soon be released
• Angiotensin-receptor blockers, selective
  aldosterone antagonists
• Equivalent BP control not fully achieved
• Step-up agents ? somewhat artificial regimen for
  ACE group ? high BP in ACE group?
• Mean BP well below 140/90 mm Hg in all groups
• Did not include low-risk individuals nor a
  wash-out period
• Information on previous AHT meds not collected

45
Conclusions

• As 1st-step agents, ACEI, CCB, and a-blockers add
  no value over and above diuretics in preventing
  CHD or other major forms of CVD
• Less effective in preventing HF
• More expensive than diuretics

46
Conclusions

• Lowering high BP is of fundamental importance in
  reducing CVD risk
• How BP is lowered does matter
• Diuretics should remain the preferred 1st step
  drugs for treatment of hypertension
• Diuretics should be a cornerstone in the arsenal
  for care of hypertensive patients.

47
Other Remarks

• Surprising ALLHAT findings
• ACEI not the best in preventing CV events
• CCB not the worst in terms of CHD and deaths
• Expectations derived from preclinical studies,
  extrapolation from surrogate outcomes, and
  case-control and other observational studies
• Results from randomized, double-blind, clinical
  endpoint trials needed whenever possible as basis
  for therapeutic decisions