Transcript Lecture 9

Treatment and Control Groups in
Clinical Trials
ICH Guidance on Investigational
Treatment (Pertains to Drugs)
• Name of product, dosing, dosing schedules,
route of administration, treatment duration
• Medications permitted, including rescue
medication
• Procedures for monitoring compliance.
Defining the Treatment Group
• Early studies of genotypic resistance testing
– Is the intervention a genotypic resistance test or
a genotypic resistance test + expert advice?
AIDS 2000; 14:F83-F93.
• Trials evaluating diagnostic devices for heart
failure
– Is the intervention the wireless pulmonary artery
hemodynamic monitoring or the hemodynamic
monitoring + reminders/consultations with given
to physicians and nurses? Lancet 2011; 377:658-676.
There is confounding of the diagnostic information and
advice/reminders/consultations.
Broader Issues – What is the
Question?
• Explanatory or pragmatic (practical) trial
• Comparison of treatment strategies or
policies
– Lifestyle interventions (e.g., MRFIT)
– Timing of use of treatment (e.g., ART, dialysis)
– Immediate versus deferred treatment (e.g.,
home care following MI and hospital care if
needed, aggressive versus less aggressive
treatment for glaucoma)
Control Group: Definition
• The reference group or standard treatment
against which a new treatment is compared
• The basis for comparison in a clinical trial
• Anchor for comparison
Trial reports should have many phrases like “as
compared to…”, “versus”, “different than”, “similar
to” – all with reference to the control group.
Types of Controls
• Concurrent
– Randomized
•
•
•
•
•
Placebo (in lieu of treatment or on top of standard of care)
No treatment
Deferral of active treatment
Take away an active treatment
Active
–
–
–
–
–
Single treatment (could be a different dose of same drug)
Investigator/patient’s choice of treatment
Optimized management or standard of care
Dynamic treatment regime
Usual care
• In some situations multiple control groups are of interest (e.g., multiple active
arms or an active arm and a placebo arm)
– Non-Randomized
• Historical (e.g., patient registry)
• Combination of randomized and non-randomized (concurrent or
historical)
One of Muench’s Postulate
“Nothing improves the performance of
therapy like the weakness of controls in its
appraisal.”
Some Recent Examples
• Cox 2 Inhibitor Studies
– VIGOR – rofecoxib (Vioxx) versus naproxen
– APPROVe – rofecoxib (Vioxx) versus placebo
– MEDAL – etoricoxib versus diclofenac (should the
comparator have been naproxen?)
• Vitamin K Antagonists for Patients Atrial Fibrillation
– Dabigatran was found to be superior to wafarin. Should
other new agents be compared against dabigatran
N Eng J Med, January 25, 2007 and October 27, 2011
Review of Industry Sponsored Studies
• 577 trials
– 187 placebo controlled
– 285 active controlled
– 105 dose comparisons
• 82% with single industry sponsor
• 18 head-to-head comparison trials of active
interventions owned by different companies.
Lathyris D et al, Eur J Clin Invest 2010
Placebos
Placebo (def.) - A medication prescribed more for
the mental relief of the patient than for its actual
effect on the disorder; something tending to
soothe (Webster’s Dictionary)
Note: Definition pertains to drug studies but use of
“placebos” is not confined to drug trials (e.g.,
acupuncture study)
A pharmacologically inactive agent to maintain
blinding in a clinical trial (no specific action on the
patient’s symptoms or disease)
Angina Pectoris
and the Placebo Effect
(Benson, H., N Eng J Med, 1979)
Subjective improvement for five 82.4 ± 9.7%
inactive treatments used prior to
1960 (13 studies, 1187 patients)
Other findings:
• Increased exercise tolerance
• Reduced nitroglycerin usage
• Improved ECG results
How Big is the Placebo Effect?
• Systematic review of trials that included a
placebo and no treatment group
• 114 trials
• No significant difference between placebo
and no treatment except in trials with
subjective continuous outcomes and in trials
for pain treatment.
N Engl J Med 2001; 344:1594-1602.
Treatment of Mild Hypertension Trial
• Selected self-reported side-effects (placebo
versus active treatment)
– Weakness 18% versus 16%
– Headaches 34% versus 22%
– Muscle pain 33% versus 26%
Arch Intern Med 1991; 151: 1413-1423.
Example
Effect of Antihistamine on Colds Under
One Day Duration. Br Med J, 1950.
After 2nd day of treatment:
Antihistamine
Cured
Cured/
Improved
13.4%
68.2%
Second Day of Treatment:
Cured
Cured/
Improved
Antihistamine
13.4%
68.2%
Placebo
13.9%
64.7%
Example illustrates that placebos control for:
1) Suggestion (placebo effect)
2) Changes in course of disease (also regression to
the mean)
The latter, but not the former, could be controlled for with an open-label
randomized study with a no-treatment control.
Main Reasons for Using Placebos
• Facilitates blinding
• Controls for the placebo effect
Placebo Effect
Subjective changes as well as objective
physiological changes (beneficial and toxic)
produced by placebo (not limited to psychological
responses)
See Beecher HK, JAMA 1955.
Can the Placebo be the Cure?
Science, April 9, 1999
“Merck was struck by the curse of the
placebo effect…patients who had received
a dummy pill had done unexpectedly well.
…it highlights a chronic problem for
psychopharmacology – the placebo effect”
Example of Impact
of Placebo Responders
on Results of a Trial
A Crossover Experiment of Four Treatments
Treatment
Order
1
A
B
C
D
Period
2
3
B
C
A
D
D
A
C
B
4
D
C
B
A
Reference: Jellinek, Biometrics Bulletin
A=a+b+c
B=a+c
C=a+b
D = placebo
Response variable = fraction of headaches relieved
A
B
C
D
0.84
0.80
0.80
0.52
A Comparison of Response to
Treatments A, B and C by the
Response to D (Placebo)
Patients Who
While on Placebo Received
Some Relief
A
0.82
B
0.87
C
0.82
120/199 Subjects
Patients Who
While on Placebo Received
No Relief
A
0.88
B
0.67
C
0.77
79/199 Subjects
Placebo Effect Summary
1. Placebos can be very effective particularly as judged
by subjective response variables
2. Placebos control both for suggestion and
spontaneous changes in course of disease – usually
difficult to disentangle
3. The placebo effect can be short-lasting (angina
studies)
4. Removal of “placebo responders” before
randomization (more later on such enriched designs)
– May make it easier to detect treatment differences; and
– Impacts generalizability (sacrifice some external validity)
Considerations in Using a
No Treatment/Placebo Control
• No standard therapy with established efficacy
• Cost/availability of standard therapy
• Size/clarity of results
• Toxicity of test treatment
• Risk to patients
• Informed consent
Considerations
for Use of Placebo Controls
Is there effective treatment?
Yes
Is active control equivalence
study informative?
No
Does treatment affect survival
or irreversible morbidity in
population to be studied?
Can add-on study provide
information?
No
Yes
Placebo control acceptable
Active control acceptable
Placebo control acceptable
Yes
No
No
Yes
Add-on study
Yes
Short-term
placebo-controlled study
Can short-term study that is
ethically acceptable provide
No
needed evidence?
Is effective treatment accepted
uniformly as standard treatment?
Yes
Might new treatment prove
Yes
superior to active control?
No
Placebo-controlled trial
where doubts exist
Active control (superiority) study
From: Ellenberg and Temple, Annals Int Med, 2000;133:464-70.
International Conference on Harmonisation
of Technical Requirements for Registration
of Pharmaceuticals for Human Use
Advantages of placebo-controlled trials
• Ability to demonstrate efficacy credibly
• Measures “absolute” effectiveness and safety
• Efficiency
• Minimizes effect of subject and investigator expectations
EID: Choice of control group in clinical trials
http://www.itpma.org/ich1.html
Controversial Issues Concerning Use
of Placebos
• Use of placebo in short-term studies for conditions
where there are established treatments with
known long-term benefits, e.g., what is short-term?
how much risk is tolerable?
• Withdrawal of active treatment and random
assignment to new drug or placebo, e.g.,
psychiatric trials
• Definition of “standard of care” varies from country
to country.
Perinatal Transmission of HIV
• PACTG 076 established that efficacy of
zidovudine (AZT) in preventing transmission
from mother to child (67.5% relative
reduction)
• AZT – antepartum (oral), intrapartum (IV),
plus AZT for newborn
• Are placebo controlled trials of simpler, less
costly regimens in other countries ethical?
In many active controlled trials,
placebos are used to facilitate
blinding.
Patients receive exactly the same treatment
that they would receive if they are not in the
trial, e.g., standard of care, optimized
background treatment, usual care.
Treatment of Mild Hypertension Study
(TOMHS)
Weight Loss + Na Reduction +
Alcohol Reduction
and
(1)
Placebo
(2)
Acebutolol
(400 mg)
(3)
Amlodipine
(5 mg)
(4)
Chlorthalidone
(15 mg)
(5)
Doxazosin
(2 mg)
(6)
Enalapril
(5 mg)
“Add On” or “Augmentation” Study
AZT
AZT
+
+
3TC
vs.
3TC
+
+
Indinavir
Placebo
Hammer et al., NEJM 1997; 337:725-33.
There are a large number of trials like this –
new treatment added to “optimal” background or standard of care.
CONVINCE Clinical Trial
Hypertension Trial with Active Control that was
Patient/Clinician Choice
Randomization
Verapamil
+
Placebo for control (either
hydrochlorothiazide or atenolol)
Hydrochlorothiazide/
Atenolol
(Clinician’s choice;
pre-specified)
+
Placebo for
Verapamil
Once Daily versus Twice Daily Treatment
Once Daily
Treatment
Twice Daily
Treatment
A = slow release
(200 mg)
A = conventional tablet
(100 mg)
B = placebo
B = conventional tablet
(100 mg)
“Take Away” or Withdrawal Studies
Patients infected with HIV taking prophylaxis for
M. Avium with high CD4+ counts – Randomize:
Withdraw OI
Prophylaxis
(Placebo)
vs.
Continue
Prophylaxis
e.g., Azithromycin vs. Placebo for M. Avium Complex (see N Engl J
Med 2000; 342:1085-1092)
Focus of Comparison:
Treatment vs Regimen vs Strategy
• Treatment – comparison of a specific treatment at
a specific dose for a specific duration (e.g., many
pivotal trials done for approval)
• Regimen – comparison of one or more treatments
as part of a regimen where dose can be varied
and/or components of the regimen varied
• Strategy – includes plans for “second-line” and
auxiliary treatments
2010 National Research Council report:
The Prevention and Treatment of Missing Data in Clinical Trials
Immediate Versus Deferred
Treatment: Concorde Study
Randomization
If AIDS/ARC or CD4+
declines to
<500 cells/mm3
AZT
Placebo
Open-label
AZT
Open-label
AZT
Lancet 1994; 343:871-881.
Dynamic Treatment Regimes
• New treatments are added/subtracted based
on the history of response to past treatments
• Examples:
– ddI versus ddC “switchover”
– Invasive versus conservative (deferred
catheterization, angiography and
revascularization) rmanagement of MI
(VANQUISH)
– STAR*D for the management of depression
(multiple randomizations with ability to opt out)
(see N Engl J Med 2006;354:1211-1242.
Lavori P, Cont Clin Trials 2004; 1:9-20
Multiple Risk Factor Intervention Trial
(MRFIT): Usual Care Control
Randomization
Special Intervention (SI) aimed at
lowering blood cholesterol
and blood pressure
and smoking cessation
Usual Care (UC)
What is usual care (considering variations
in clinical practice, temporal trends,…?
JAMA 1982; 248:1465-1477
In MRFIT “Usual Care” Was Pretty Good!
∆ DBP (mm Hg)
∆ cholesterol (mg/dl)
% quitting smoking
Risk Factor Changes
After 6 Years
SI
UC
-10.5
-7.4
-12.3
-6.4
46
29
MRFIT Illustrates Challenges Defining
“Usual Care”
• Tension between control over experimental
conditions versus relevance to clinical care – how
strictly should control treatment be specified?
(explanatory versus pragmatic approach)
• Best practices may be under-utilized in the “realworld”
• Evidence base to guide usual care may not be
optimal and may change over a long-term study.
See also Dawson L et al, PloS Med 2009;6:e1000111
Dose Comparison/Escalation
Low vs. Intermediate vs. High Dose vs. Placebo
or
Low Dose vs.
Placebo
Int. Dose (If “Poor” response to Low Dose)
Hi Dose (If “Poor” response to Int. Dose)
Circular Motion
(Bradford Hill)
Cannot compare dose levels from a trial
designed to compare two forms of
management.
If dose is determined by response of
patient, then responses of patients at
different dose levels cannot be compared.
N Engl J Med 1953; 247:113-119.
Often Cited Reasons
for Uncontrolled Studies
1. Unnecessary for large effects
2. Controlled studies are more difficult to implement
3. Availability of patients
4. Ethical reasons
– no control treatment available
– untreated patients at high risk of death or
serious illness
5. Historical data for comparison is available
Historical Controls Could Result in An
Effective Treatment Being Abandoned:
REMATCH Trial in Advanced HF
Control
Medical Treatment
2-Year Death
Rate (%)
Treated
LVAD
2- Year Death
Rate (%)
Design assumptions
75%
50%
Results of trial
92%
77%
NEJM 2001 15:1435-1443.
Summary
Considerations in Choice of Experimental and
Control Treatments
• Acceptability by patients and clinicians
• Indifference/doubt concerning relative
efficacy/safety
• Some basis for thinking that there could be
a difference of clinical/public health
importance
Meinert gives the following requirements
for experimental and control treatments:
• Must be distinguishable
• Medically justifiable
• Ethically OK
• Both must be acceptable to patients and investigators
• Reasonable doubt concerning efficacy
• Should be reason to believe benefits outweigh risks
• Method of administration should be as similar as
possible to real-world use
Ref. Clinical Trials: Design, Conduct and Analysis