1. Declaration of Helsinki and Active Control Trials
Robert J. Temple, M.D.
April 11, 2002
Declaration of Helsinki and Active Control Trials
Robert J. Temple, M.D.
Associate Director for Medical Policy
Center for Drug Evaluation and Research
U.S. Food and Drug Administration
April 24, 2002
Declaration of Helsinki and Active Control Trials

2. Placebo Controls/Active Controls
Robert J. Temple, M.D.
April 11, 2002
Placebo Controls/Active Controls
Arguments about use of placebos are sometimes presented as one of conflict between science and ethics, a case where one sometimes weighs the two on a common scale
That is incorrect. There are two issues, but they are distinct: one ethical, one inferential
1. Ethical - when is it O.K. to deny patients known effective therapy by randomizing some to a placebo
2. Inferential - when can an active control equivalence study (the alternative to a placebo) demonstrate effectiveness
Declaration of Helsinki and Active Control Trials

3. Active Control Trials
Active control trials (comparing new and standard therapy and showing no difference between them) are usually proposed as a way to demonstrate effectiveness because a placebo-controlled trial is considered unethical or undesirable. It seems a simple and logical alternative but in fact the design and interpretation of such trials is often very difficult. In many cases the needed conditions for utilizing such a trial as evidence of effectiveness cannot be met and only a placebo-controlled trial will be interpretable (but may not be possible on ethical grounds)
In what follows I will discuss the ethics of placebo-controlled trials and the necessary conditions for use of active control trials

4. The Ethical Issue: Need to Use Effective Available Therapy
If there is known effective therapy for a condition, when, if ever, is it ethical to deny this treatment to some patients in a clinical trial?
Doesn’t it depend on what condition is being treated and the consequences of delayed or omitted treatment?

5. Declaration of Helsinki, 1975
In any medical study, every patient - including those of a control, if any - should be assured of the best proven diagnostic and therapeutic method
What did the 1975 Declaration mean?

6. Ethical Issue
Some (e.g., Rothman, NEJM, BMJ) have contended that the 1975 Declaration must be read literally, and that the 2000 revision strengthens their view, and therefore the condition being treated is irrelevant. Thus:
• No placebo-controlled trials in baldness (Rogaine)
• No placebo-controlled trials in seasonal allergic rhinitis
• No placebo-controlled trials in headache
• No placebo-controlled trials of any duration in insomnia, anxiety, outpatient depression, OCD
Read literally, however, Declaration bars any trial (even active comparison) if there is existing therapy (people receiving the test drug don’t receive “best proven” treatment

7. What Did Declaration (1975) Mean?
In presenting the 1975 change, the WMA did not say a word about placebo controlled trials. It said the change was intended to reinforce the idea that the physician-patient relationship “must be respected just as it would be in a purely therapeutic situation not involving research objectives”

8. How the 1975 Declaration Affected Placebo-Controlled Trials
It had little or no effect. Despite Rothman and Michels (and a few others), placebo-controlled RCTs were performed, despite existing therapy, by the 1000’s, were published in journals that would not publish an unethical trial. The people performing and publishing the trials considered them ethical and compatible with the Declaration
Then the Declaration was changed

9. Declaration of Helsinki, 2000
The benefits, risks, burdens, and effectiveness of a new method should be tested against those of the best current prophylactic, diagnostic, and therapeutic methods. This does not exclude the use of placebo or no treatment, in studies where no proven prophylactic, diagnostic, or therapeutic method exist.

10. Declaration of Helsinki, 2000
The first sentence could be read as a scientific statement (know how new Rx compared with old), perhaps supporting 3-arm (test drug, control, placebo) studies. The second sentence makes it clear that was not the intent

11. What is Ethical Impediment to Placebo-Controlled Trials in Symptomatic Conditions?
There never has been a coherent explanation by Rothman, the WMA, or others of why an informed patient cannot volunteer to defer treatment of a symptomatic condition or allow a brief deferral of treatment of elevated blood sugar, blood pressure, or cholesterol (but can, in contrast, take a new, untested drug instead of established therapy)
No impact on health
Easily explained; consequences understandable
Context of trials not misleading (doesn’t look like it is treatment); i.e., should be no “therapeutic misconception”
Easy escape if treatment unsatisfactory
Contrast with clear Declaration of Helsinki statement that healthy volunteers may participate in research

12. Similar to choices people make every day to treat, or not to treat, symptoms
Pain treatment (headache, dental, arthritis)
Many psychiatric conditions (anxiety, OCD, depression)
Seasonal allergies
Erectile dysfunction
Baldness
BPH symptoms
Insomnia
Heartburn, dyspepsia
Angina
Similarly, laboratory and clinical abnormalities are often untreated for brief periods (e.g., 2-4 weeks)
Mild hypertension
Elevated cholesterol
Elevated blood sugar

13. It’s the Consequences of Not Receiving Standard Therapy
The recently published FDA guidance, Choice of Control Group and Related Issues, a guidance developed by ICH (ICH E-10) takes a very different position
It clearly distinguishes between current treatments that prevent serious harm and those that treat symptoms. Whether a placebo-controlled trial is ethical depends on the consequences of non-treatment

14. ICH E-10 Guidance
The principal issue [in use of placebos] is the ethical one. There is no issue when no effective therapy exists. The question is when is it acceptable not to give existing therapy and randomize to drug or placebo.
“In cases where an available treatment is known to prevent serious harm, such as death or irreversible morbidity in the study population, it is generally inappropriate to use a placebo control.” (Generally, because if treatment is so toxic that people refuse it, a placebo may still be possible.)

15. ICH E-10 Guidance (cont’d)
“In other situations, where there is no serious harm, it is generally considered ethical to ask patients to participate in a placebo-controlled trial, even if they may experience discomfort as a result, provided the setting is non-coercive and patients are fully informed about available therapies and the consequences of delaying treatment… Whether a particular placebo-controlled trial of a new agent will be acceptable to subjects and investigators when there is known effective therapy is a matter of investigator, patient, and IRB judgement, and acceptability may differ among ICH regions. Acceptability could depend on the specific trial design and population chosen.”

16. There Are Many Situations in Which Placebo is Unacceptable
All agree that a placebo-controlled trial is unacceptable when there is adequate evidence that available treatment can increase survival or prevent irreversible morbidity in the intended study population over the course of the proposed study. Examples:
1. Post-infarction trials of thrombolytics, beta blockers, aspirin, ACEI’s (in patients with ventricular dysfunction)
2. Antibiotic prophylaxis in “dirty surgery”
3. Initial treatment of childhood ALL or testicular cancer
4. Long-term trial of moderate to severe hypertension
5. CHF of virtually any severity
NB: Active control equivalence study in these cases may not be interpretable

17. There May Be Close Cases
In some cases there can be debates about whether the risks of non-treatment of acceptable. For example:
Trials involving withdrawing people from antipsychotics to enter the study (rarely done; usually new patients)
Anti-emetics in severely emetogenic chemotherapy
Duration of non-treatment of BS, BP, cholesterol

18. Metaanalysis of 17 Placebo-Controlled Trials [Beasley et al
Metaanalysis of 17 Placebo-Controlled Trials [Beasley et al. BMJ (1991); 303:685-92]
Prozac Tricyclics Placebo n
Suicidal Acts 6 (0.3%) 3 (0.4%) 1 (0.2%)
New Substantial 14 (1.2%) 15 (3.6%) 10 (2.6%)
Suicidal Ideation
Worsening Suicidal 262 (15.3%) 117 (16.3%) 100 (17.9%)
Ideation
All studies had placebo lead-in: 3 suicidal acts; 1 fatal

19. Recent Activity, U.S. and Other
Guidance, March 2001 on interpreting reference to Declaration of Helsinki in 21 CFR (conforms to ethical principles contained in the Declaration of Helsinki)
It is the 1989 version of Declaration of Helsinki that is referred to

20. Recent Activity, U.S. and Other
EMEA/CPMP position paper of June 28, 2001
notes need for placebo in many situations to obtain reliable evidence
acceptable if no added risk of irreversible harm
WMA press release 8 Oct 2001.
Clarification: placebo controls can be acceptable, even if proven therapy exists, if
there are compelling methodological reasons
in minor conditions where patients receiving placebo “would not be subject to any additional risk of serious or irreversible harm”

21. Why Placebo Controls May Be Needed, Even If There is Existing Treatment
The alternative to a placebo-controlled trial (or other trial showing a difference between treatments, such as a dose-response study or a study showing superiority to an active control) is an active control “equivalence” trial, showing that the new drug is similar to a drug with known effectiveness
But equivalence trials, in many situations, are uninformative or misleading

22. Problems with “Equivalence” Trials
Studies that are designed to show “equivalence” or “non-inferiority” have three principal problems.
1. There is a critical assumption: that the trial had “assay sensitivity”
2. There is no standardized test for “statistically significant similarity,” so that it is critical to choose a difference between treatments that, if ruled out by the study, will demonstrate effectiveness. Choosing this difference is difficult
3. Lack of incentives to study excellence

23. Assay Sensitivity
There is an inherent, usually unstated, assumption in all such trials, namely that the active control was effective in the particular study in question, i.e., that the trial was capable of distinguishing effective from ineffective drugs
The assumption:
• is not necessarily true for all effective drugs
• is not directly testable in the data collected
• in essence, causes an active control study to have elements of a historically controlled study

24. Evidence of Efficacy Two distinct approaches: 1. Difference-showing
Superiority of test drug to control (placebo, active, lower dose) demonstrates drug effect (and assay sensitivity)
2. Equivalence or non-inferiority in an active control study
Showing similarity to a known effective therapy (active control) and attributing the efficacy of the active control drug to the new drug, thereby demonstrating drug effect
Non-inferiority trials show that the new drug is not worse than the control by a defined amount, that amount being no larger than the effect the active control would be expected to have in the study (the non-inferiority margin, M)

25. The Logic Is Not The Problem
Showing equivalence to a known active drug would be a sensible way to demonstrate effectiveness
But you can’t really show equivalence (except by being superior), so we seek Non-Inferiority,
a misnomer
Really it is showing inferiority of no more than a specified margin M, or delta
C-T<M
So it’s really a “not-too-much-inferiority” trial
[Old, naïve way (but still seen in current publications) was to compare C and T, find “no significant difference” and declare victory. Here, at least, we’re past that]

26. Non-Inferiority Trials
Specify as a null hypothesis that the difference between active control (C) and new drug (T) is greater than some amount M (the margin), M being no greater than the effect of C in that study; if the null hypothesis is not rejected, T has no effect at all. That is:
Ho: C-T > M
Ha: C-T < M
Usually, in analogy with placebo controlled trials, the 97 1/2% CI (one-sided) for C-T must be < M, demonstrating a “statistically significant” effect of T.

27. Assay Sensitivity
Interpretation of the study depends completely on M, the effect of C that is presumed for the new study. But M is not measured; it must be deduced from historical experience (generally placebo-controlled trials with C)
Equivalence or non-inferiority trials always raise a question:
Did the active control drug have an effect of the size expected in the trial carried out? If it did not, equivalence or non-inferiority by the expected effect is meaningless. The equivalent or non-inferior drug could have no effect at all.

28. Fundamental Logic of Trials
Superiority = Efficacy
(if control > placebo)
Non-inferiority  Efficacy
(unless assay sensitivity is present)

29. Determining Assay Sensitivity
To conclude a trial had assay sensitivity, you need a combination of historical information and information about the new trial.
Historical evidence of sensitivity to drug effects
A historically based conclusion that appropriately designed, sized, and conducted trials in a particular disease, with a specific active drug (or group of related drugs) reliably show an effect of at least some defined size on a particular endpoint. Usually established by showing that appropriately sized (powered) and well-conducted trials in a specified population regularly distinguish the active drug(s) from placebo for particular endpoints
Sensitivity to drug effects is an abstract conclusion about well-designed trials of a drug in a particular disease. Assay Sensitivity is a conclusion about a particular trial

30. Historical Evidence of Sensitivity to Drug Effects: a Precondition for Assay Sensitivity
Examples of situations in which studies of current drugs lack sensitivity to drug effects:
• Depression
• Anxiety
• Dementia
• Symptomatic congestive heart failure
• Seasonal allergies
• Gastroesophageal reflux disease
• Post-infarction beta blockade
• Post-infarction aspirin

31. Historical Evidence of Sensitivity to Drug Effects
Conclusion of HESDE applies only to trials of a particular design (patient population, selection criteria, endpoints, dose, use of washout periods) . Changes in these can affect the effect size of the active control and, therefore, the appropriate margin, or completely undermine assay sensitivity

32. Assay Sensitivity and Study Quality
If sensitivity to drug effects exists for a therapeutic class, assay sensitivity in a particular study can still be undermined by a variety of study conduct factors that “bias toward the null,” i.e., obscure true differences between treatments
These factors include:

33. Assay Sensitivity and Study Quality
Poor compliance
Non-protocol crossovers
Spontaneous improvement in the population
A poorly responsive population
Use of concomitant medication that interferes with test and control treatment or that reduces extent of potential response
Poor diagnostic criteria (patients lack the disease to be studied)
Inappropriate (insensitive) measures of drug effect
Poor quality of measurements
Mixing up the treatments

34. Assay Sensitivity and Study Quality
These factors, in general, have only small (or no) effects on variance (width of CI) but can reduce or obliterate A-T differences, leading to false conclusion of non-inferiority
Note: Some analytic approaches that are “conservative” in a difference-showing trial are not in a non-inferiority trial; for example, an intent-to-treat approach reduces A-T and is not conservative

35. The Non-Inferiority Margin
M has two properties of interest
1. It must be no larger than the entire effect of the control drug (C) in this study. Call this M1
2. It cannot be larger than a clinical difference (degree of inferiority) you are willing to accept. This is M2

36. Find M1, First
To design a non-inferiority study, find M1, the effect you are sure the control agent had in the study. Then if
C-T<M1,
the test drug has at least some effect
Then decide how much of M1 you need to preserve, say at least 50%. In that case, M2 will be the reference for the null (really non-inferiority) hypothesis and the study will need to show
C-T<M2

37. M1 M2 M0 -2  Difference in Effect (ES - ET ) 1 2 3 4 5 -1 2 1
Robert J. Temple, M.D.
April 11, 2002 1 2 3 4 5 -1 -2 2 1 M0 M1 M2
 Difference
in Effect
(ES - ET )
Declaration of Helsinki and Active Control Trials

38. Illustrations of Lack of HESDE
Depression and other psychiatric conditions
Symptomatic CHF
Post-operative nausea and vomiting

41. Historical Evidence of Sensitivity to Drug Effects: a Precondition for Assay Sensitivity
Examples of situations in which studies of current drugs lack sensitivity to drug effects:
• Depression
• Anxiety
• Dementia
• Symptomatic congestive heart failure
• Seasonal allergies
• Gastroesophageal reflux disease
• Post-infarction beta blockade
• Post-infarction aspirin

42. Conclusions
1. Placebo controls are ethical, even when there is existing effective therapy, if subjects on placebo will not be harmed
2. Active control non-inferiority trials are informative only if you can be sure the trial had assay sensitivity and could have detected an effect vs. placebo (had there been a placebo) of M1, the presumed effect of the active control
3. Assay sensitivity often cannot be presumed