Sham Controls in Medical Device Trials: Perspective
Sham Controls in Medical Device Trials: Perspective
Sham Controls in Medical Device Trials: Perspective
positive response to sham acu- ward if two groups of patients risks were weighed and managed.
puncture, while only 22% had a were randomly assigned to a car- There were risks associated with
positive response to oral pharma- diac catheterization procedure, as the sham control, which includ-
cologic placebos.5 This research was done for renal-artery dener- ed a femoral-artery puncture and
shows not only an astonishingly vation, such a study has yet to be renal angiography. However, the
high response rate for sham pro- performed, and the important finding that the procedure lacks
cedures, but also a significantly question of PCI’s actual clinical any apparent benefit will spare
higher response rate for placebo benefit therefore remains unan- many patients from undergoing
physical interventions than for swered. a risky procedure that apparently
placebo drugs. These results high- Subjecting patients to sham has only placebo value. There are
light the importance of devising procedures is not without risk, clear benefits of preventing inef-
a control that will sufficiently and it gives rise to ethical con- fective procedures and devices
distinguish the specific effect at- cerns about “unnecessary” inva- from gaining widespread use,
tributable to the placebo. sive procedures that will have no which means that true therapeu-
Not all device trials necessar- actual therapeutic effect. I believe, tic benefit should be established
ily require nontherapeutic con- however, that the examples above before FDA approval.
trols. For example, after a thera- show that sham interventions are The SYMPLICITY trial thus
peutic benefit beyond the placebo ethical when the benefits of in- adds to mounting evidence that
effect was established, subsequent formation from a sham-control medical procedures can have a
iterations of a device would not trial exceed the risks of using an substantial placebo effect. This
need to be compared with a intervention not shown to be knowledge may require Congress
sham control. In addition, trials more therapeutic than a sham. to articulate a clear standard for
with only objective end points, Moreover, the risk associated with establishing true therapeutic ben-
such as mortality, do not need a performing unnecessary proce- efit for FDA approval, to ensure
nontherapeutic control. Interven- dures should be weighed against that all devices we provide to our
tional studies that would most the risk of mistaking a placebo patients are safe and effective.
appropriately be conducted as effect for therapeutic benefit and Disclosure forms provided by the author
blinded RCTs include early studies therefore subjecting thousands are available with the full text of this article
of a new technology and studies or millions of patients to a pro- at NEJM.org.
whose primary outcome measure cedure that actually does them
is susceptible to a placebo effect, no good. In a controlled trial, From the University of California, San Fran-
cisco, Medical Center, San Francisco.
such as pain. patients are informed of and
For example, percutaneous cor- consent to the risks; when an in-
1. Dhruva SS, Bero LA, Redberg RF. Strength
onary intervention (PCI), a widely effective procedure is accepted of study evidence examined by the FDA in
used procedure for treating sta- into practice, however, patients premarket approval of cardiovascular devices.
ble coronary artery disease, has who subsequently undergo it most JAMA 2009;302:2679-85.
2. Bhatt DL, Kandzari DE, O’Neill WW, et al.
never been investigated in a blind- certainly have not knowingly con- A controlled trial of renal denervation for re-
ed trial. Some nonblinded RCTs sented to an ineffectual proce- sistant hypertension. N Engl J Med 2014;370:
have shown that PCI has a bene- dure. Without careful use of non- 1393-401.
3. Gray DT, Hollingworth W, Onwudiwe N,
ficial effect on anginal symp- therapeutic controls, we may be Deyo RA, Jarvik JG. Thoracic and lumbar ver-
toms, but there appears to be no subjecting millions of Americans tebroplasties performed in US Medicare en-
difference between PCI and med- to harm from risky, invasive pro- rollees, 2001-2005. JAMA 2007;298:1760-2.
4. Buchbinder R, Osborne RH, Ebeling PR,
ical therapy in rates of the objec- cedures without benefit. Ethical et al. A randomized trial of vertebroplasty
tive end points of nonfatal myo- concerns regarding a placebo for painful osteoporotic vertebral fractures.
cardial infarction and death due group should, of course, be ac- N Engl J Med 2009;361:557-68.
5. Meissner K, Fässler M, Rücker G, et al.
to cardiac causes. It is possible, knowledged and addressed by in- Differential effectiveness of placebo treat-
therefore, that the perceived symp- stitutional review boards and ments: a systematic review of migraine pro-
tomatic benefit is actually a pla- through informed consent, as they phylaxis. JAMA Intern Med 2013;173:1941-
51.
cebo effect and not attributable are in drug trials and have been
to PCI. Although a blinded trial in the examples above. DOI: 10.1056/NEJMp1406388
would be relatively straightfor- In the SYMPLICITY trial, the Copyright © 2014 Massachusetts Medical Society.