Placebo Control Arms Are Deeply Flawed Scientific Tools

Maurie Markman, MD

There is probably nothing more controversial in clinical research than the role of placebo-control studies and whether they are necessary.

From a distance, it is easy to appreciate the value of comparing a prospectively defined outcome of an investigative treatment with an outcome involving individuals who not only received an inactive therapy but also were unaware of it. Under these circumstances, the impact of a given therapeutic approach can be evaluated without the variable of a potentially biologically or clinically active control or the psychological influence of patients knowing they were not receiving a strategy that might bring about a hoped-for outcome.

Considerable evidence demonstrates the impact of positive or negative patient expectations on trial outcomes for pharmaceutical agents used for certain clinically relevant medical conditions (eg, depression, anxiety, migraines, asthma).^1,2 The results of randomized placebo-controlled trials have revealed that control study arms often perform as well in achieving measurable beneficial effects as active drugs in such settings. Further, there is strong evidence that psychological factors may also influence the objective measurements associated with medical devices.³

These experiences have provided strong support for the argument that there may be a critical role for a placebo study arm in a randomized phase III trial where the primary endpoint may be influenced by patient expectations. However, particularly in the oncology arena, there are several serious scientific, ethical, and pragmatic concerns associated with ongoing reliance on placebo-controlled studies. These are highlighted below. First, and rarely discussed within the peer-reviewed medical literature, is the fact that trials including the use of placebos often do not provide information about the manufacture or detailed composition of the compounds serving this role.⁴ For example, how does one know a so-called nonspecific antibody employed as a control in a cancer vaccine study is completely unrecognized by the immune system in comparison with the “active” agent? In the absence of such information, how can we be sure this control antibody will not have a positive or even a negative effect, rather than simply no effect, on the study endpoints (eg, progression-free, symptom- free, and overall survival)?

Second, we must acknowledge the ethical issue of randomizing a patient to a placebo to be evaluated against a potentially clinically active anticancer strategy. Regulatory requirements may stipulate that a novel antineoplastic be compared with a previously approved agent for a specific indication (eg, third- or fourth-line therapy of a refractory cancer). It is more than theoretically possible that in the absence of such a comparator that fulfills this requirement a placebo control may be required to serve as the study control.

Unfortunately, this rather rigid perspective ignores the common observation that individual patients may achieve a meaningful degree of clinical benefit from commercially available, unindicated antineoplastic therapeutic regimens, even if these do not objectively quantify or satisfy the regulatory criteria for approved agents. However, from an ethical perspective, a strong argument can be made that it would be far more appropriate for the study control to be “physician/ patient choice” to include strategies employed in similar situations in routine, real-world clinical practice.

Third, one must consider the remarkable costs—not only financial, but also time and effort—associated with the use of placebos in randomized clinical trials of cancer agents. In a placebo-controlled study, the placebo needs to be manufactured, distributed, and administered, and trial sponsors need to fund the expenses of patients being seen in a setting where a third-party payer is appropriately unlikely to agree to pay for physician and nurse time or imaging studies whose purpose has nothing to do with benefiting patients. Patients, and likely caregivers, may be required to make visits to physicians’ offices for infusions or oral administrations of inert substances that have no potential to favorably impact clinical outcomes.

Let’s conclude with a brief inquiry into the objective value of including a placebo control in cancer clinical trials. The hypothesis supporting use of this approach is that patients and physicians will not know whether the individual is receiving an active compound or placebo and, therefore, subjective psychological associations with receipt of therapy will be eliminated or at least substantially reduced.

Just how realistic is this concept? For example, when a known adverse effect of an active agent is experienced, do patients and physicians truly not make the connection or do they at least strongly suspect an active drug is being administered? How difficult would it be to correctly guess that a patient with no prior history of elevated blood pressure was receiving bevacizumab (Avastin) rather than placebo if grade 2 or 3 hypertension was observed in that individual? Reports in gynecologic oncology, for example, have noted that the incidence of grade ≥2 hypertension in patients receiving this agent has ranged as high as 49.3%.⁵

Finally, how different are the objective clinical outcomes for similar cancer studies that have employed a placebo control versus those that have not? A relevant response to this important question can be provided by 2 phase III randomized trials that examined the addition of bevacizumab to cytotoxic chemotherapy in the primary management of ovarian cancer.^6,7 Although not identical in design, both studies combined bevacizumab and chemotherapy, with bevacizumab subsequently continued as a single-agent maintenance strategy. One of these studies employed a bevacizumab placebo during chemotherapy and the maintenance phase, while the second study was unblinded and did not include a placebo control. The 2 trials’ results, published in the same issue of a high-impact medical journal, revealed quite similar results for progression-free survival in favor of the bevacizumab-containing regimen.^6,7

It is here where one must ask why the placebo control was necessary in the one study for any reason other than an existing regulatory mandate? Perhaps it is time to carefully and critically examine the regulations.

References

1. Meissner K, Fässler M, Rucker G, et al. Differential effectiveness of placebo treatments: A systemic review of migraine prophylaxis. JAMA Intern Med. 2013;173(21):1941-1951. doi: 10.1001/jamainternmed.2013.10391.
2. Wechsler ME, Kelley JM, Boyd IO, et al. Active albuterol or placebo, sham acupuncture, or no intervention in asthma. N Engl J Med. 2011;365(2):119-126. doi: 10.1056/NEJMoa1103319.
3. Redberg RF. Sham controls in medical device trials. N Engl J Med. 2014;371(10):892-893. doi: 10.1056/NEJMp1406388.
4. Golomb BA, Erickson LC, Koperski S, Sack D, Enkin M, Howick J. What’s in placebos: Who knows? Analysis of randomized controlled trials. Ann Intern Med. 2010;153(8):532-535. doi: 10.7326/0003- 4819-153-8-201010190-00010.
5. Corr BR, Breed C, Sheeder J, Weisdack S, Behbakht K. Bevacizumab induced hypertension in gynecologic cancer: Does it resolve after completion of therapy? Gynecol Oncol Reports. 2016;17:65-68. doi: 10.1016/j.gore.2016.06.002.
6. Burger RA, Brady MF, Bookman MA, et al; Gynecologic Oncology Group. Incorporation of bevacizumab in the primary treatment of ovarian cancer. N Engl J Med. 2011;365(26):2473-2483. doi: 10.1056/NEJMoa1104390.
7. Perren TJ, Swart AM, Pfisterer J, et al; ICON7 Investigators. A phase 3 trial of bevacizumab in ovarian cancer. N Engl J Med. 2011;365(26):2484-2496. doi: 10.1056/NEJMoa1103799.