There seem to be two ideas of placebo - or the treatment received by a placebo group:

  1. A "simulation" of a treatment using saline or sugar pill so that subjects in a clinical trial can be randomized, and drug can be administered in a blinded fashion.
  2. A "non-treatment" so that the relative effects of a single drug can be isolated.

For example, this real-world evidence group supposes a case where it's unethical to administer a placebo "such as in oncology studies" https://www.youtube.com/watch?v=86Rgbw_Kw8Y. But I fail to see how that can be the case except if the investigational treatment is extremely harmful.

Two examples: consider a rare form of cancer treatable only by infused chemotherapy. Treatment with chemo gives a median PFS of 3 months and OS of 6 months - in other words, the chemo isn't very good. Suppose I have a new orally administered medication which has promising anti-tumor activity in mouse models, and the phase 1A study went well, and I want to inspect anti-tumor activity of my drug as monotherapy in a randomized Phase 2 study.

If my goal is to determine the superiority of my oral drug, it seems that the appropriate choice of placebo is to randomize subjects, say 1:1, to receive either active oral drug and placebo infusion or to receive placebo oral drug and active infusion. On one hand, few people receive chemo without knowing they have it, and infusing saline for an hour is a waste of time. But I'm not exactly sure it's unethical. On the other hand, our objective is to show superiority to chemotherapy, so it seems worthless to enroll subjects who receive no anti-cancer drug at all.

  1. Which of the two definitions are the clinically accepted standard definition of placebo? (Or what variant thereof?)
  2. Are there examples of unethical placebo in an oncology study according to definition 1?
  3. Is there actually any value to quantifying option 2, i.e. comparing to a true no-treatment group when there is an established standard-of-care?

EDIT: I realize after posting this we should exclude one obvious example: say attempting surgery on a disease that's pronounced metastatic or unressectable. I see that you can't "blind" a surgery. In any case, some examples would be helpful for me.

  • 3
    $\begingroup$ The "blind" surgery is called sham surgery. $\endgroup$ Feb 25, 2022 at 18:19
  • 1
    $\begingroup$ I suppose two ethical objections may be the strongest: (1) Withholding what is now recognized as the best treatment from those who get the new drug, which may turn out to be useless. (2) Any premature administration of the new drug, which may turn out to have some bad effects in spite of "promising" phase 1 results (typically with relatively few subjects). Obviously, these objections do not prevent all clinical trials, but they are the basis for some of the strict regulations on clinical studies. $\endgroup$
    – BruceET
    Feb 25, 2022 at 18:56
  • $\begingroup$ A potential perk of a placebo controlled trial would be opening the cody at the end of the study, and informing the patient of her/his actual treatment... this would amount, at least in part, to a n-of-1 trial... $\endgroup$ Feb 28, 2022 at 11:28

1 Answer 1


A randomized controlled trial requires that there is a sincere uncertainty about the relative efficacy of the treatments under investigation. The term for this uncertainty is "equipoise."

The feeling that one group is "missing out" on a new treatment is strong, but often not justifiable; many drugs that undergo large trials do not work. An example where "missing out" represented a real harm to patients were the early trials of HIV medications. At the time, there was no treatment for HIV, and it was clearly fatal. Patients with HIV argued that placebo-controlled trials were unethical and were successful in gaining access to experimental medications, which were clearly effective and extended the lives of persons with HIV.

If a reasonable treatment exists for a disease, any new treatment can only be compared to that existing treatment. In such cases, the treatment arms could either be:

  1. Usual care versus usual care + new treatment
  2. Usual care + placebo versus usual care + new treatment

In your example, you could compare IV chemotherapy + new oral chemotherapy to IV chemotherapy alone. If the efficacy of IV chemotherapy is truly uncertain or potentially harmful, then one could compare the new oral chemotherapy to a placebo.

  • 3
    $\begingroup$ It can also be usual care vs new treatment. Both arms get some Tx. This is called a "non-inferiority trial". $\endgroup$ Feb 25, 2022 at 19:29
  • 1
    $\begingroup$ That is a reasonable group assignment, but non-inferiority refers to a statistical procedure where one only seeks to determine whether treatment A is more harmful than B (ie, A is not inferior to B on the basis of harm). $\endgroup$
    – Todd D
    Feb 25, 2022 at 19:32
  • $\begingroup$ Non-inferiority can be used to show that A is 'just as good as' B. It doesn't have to be harm. $\endgroup$ Feb 25, 2022 at 19:43
  • 2
    $\begingroup$ In the medical literature, studies that test whether A is as good as B are termed "equivalency trials." These are rarely performed due to the very large sample sizes required and a lack of interest in whether A is as good as B. Medical device studies are an exception to this rule; the Food and Drug Administration only requires that a new device be non-inferior or equivalent to an existing device. $\endgroup$
    – Todd D
    Feb 25, 2022 at 19:47
  • $\begingroup$ I'm a biostatistician, I'm familiar w/ the medical literature. A is as good as B can include A is better than B. An entity promoting a new intervention of some kind would be perfectly happy if it turns out to be better than an existing approach. They only need to show it's not worse. Non-inferiority is not so exotic. It's covered in intro clinical trials texts such as Friedman et al, Fundamentals of Clinical Trials. A pubmed search finds 104 NI trials published last year. It may help you to read this explainer. $\endgroup$ Feb 26, 2022 at 3:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.