In the August 2006 issue of
JAOA—The Journal of the American Osteopathic Association, John C. Licciardone, DO, MBA, and David P. Russo, DO, MPH,
1 published an original contribution on the topic of blinding protocols in randomized controlled trials (RCTs) for osteopathic medical research. Their
JAOA article was related to a study on chronic low back pain that was published earlier in
Spine.
2 In the
Spine article, Drs Licciardone and Russo,
2 along with five other osteopathic medical researchers—Scott T. Stoll, DO, PhD; Kimberly G. Fulda, MPH; Jeffrey C. Siu, DO; William B. Winn, DO; and John N. Swift, Jr, DO—studied the effects of osteopathic manipulative treatment (OMT) on chronic low-back pain in three groups of randomly assigned patients. The investigation had one study group that received OMT and two control groups, one that received sham manipulative treatment while the other received no manual intervention. All subjects continued to receive whatever other treatments (“cotreatments”) were currently prescribed to them for symptom control.
Study outcomes demonstrated no statistically significant difference between the two manual intervention groups (ie, OMT and sham), though subjects in both groups had better outcomes than did the nonintervention control group. Licciardone and colleagues
2 acknowledged several factors that could have contributed to the lack of treatment superiority observed in the OMT group when compared with subjects receiving the sham intervention, including (1) manual interventions were provided by predoctoral osteopathic manipulative medicine (OMM) fellows, (2) a small sample size reduced researchers' abilities to detect small to moderate treatment effects, and (3) a high rate of attrition. In addition, though the sham intervention protocols took care to avoid areas of diagnosed somatic dysfunction, they required a rather extensive series of movements, light touch, and “simulated OMT techniques” of diminished magnitude. This manual control protocol could certainly have had an effect on low back pain, especially when delivered by well-trained OMM fellows.
In their August 2006
JAOA article, Licciardone and Russo
1 analyzed another aspect of the data gathered during the study, subject expectation. Pretreatment subjects reviewed two brief passages of text that explained each of the manual interventions used. Subjects then received a brief questionnaire that inquired as to their expectations for treatment using a Likert scale response to the following statement: “I am confident that this treatment can alleviate my complaint.” Subjects had slightly lower expectations for the treatment efficacy of the sham statement than the treatment statement. In other words, subjects found the written explanation of OMT (
Figure 1) to be slightly more credible as a treatment option than was the description of sham manipulative treatment (
Figure 2). Researchers reassessed subject expectancy at 6-month follow-up and determined that credibility ratios remained basically unchanged over time from baseline measures (
P=.79). Although the authors
1 noted that such a difference in subject expectation or credibility could affect trial outcomes, they indicated that it did not appear to do so in their investigation. This conclusion was based on the fact that there was a significant credibility difference in favor of the OMT description but no significant treatment effect between the two manual intervention groups.
When paired, these articles put into bold relief many of the difficulties osteopathic researchers encounter when designing OMT trials. In these two studies,
1,2 it is possible that the difference in subject-perceived treatment credibility could have come not from the written descriptions of the two manual intervention protocols, but from patient treatment history—an unknown factor in the study. Although exclusion criteria prohibited subject use of concurrent manual therapies, treatment history and knowledge of OMT (or other manual therapies) was neither assessed nor inventoried at trial entry—much less noted in the study's inclusion or exclusion criteria. If a few subjects in the manual control group had a treatment history that included OMT, the description of sham manipulative treatment (
Figure 2) might have been unconvincing and could conceivably have been sufficient to explain the lower relative credibility ratio noted for sham manipulative treatment at trial entry and at 6-month follow-up.
Issues related to establishing appropriate control groups for OMT trials have been discussed extensively, but no firm consensus has been reached.
3,4 Perhaps it is time to look more deeply at the following concerns:
As Licciardone and Russo rightly note in their August 2006 article,
1 the gold standard for biomedical research is the double-blind RCT. This trial design was developed in the 1940s and 1950s with the goal of determining the effects of pharmacologic agents on certain infectious disease processes.
3,5 This type of study is now known as an “explanatory” or “fastidious” study,
6 as opposed to the pragmatic study—by far the more common route for studies in osteopathic medicine.
7
For explanatory studies, a linear cause-and-effect relationship is assumed between the administered agent and the therapeutic effect. That is, the drug is credited for any changes seen. Thus, any other actions affecting outcome are disregarded and are usually termed “placebo effects.”
Many articles are published each year that include some discussion or aspect of the placebo effect. In a recent search for materials published from 2003 to 2006, I was able to locate 3755 such articles cataloged by the National Library of Medicine. But the question remains: can the placebo effect indeed have an effect on the outcome of an RCT—and is that outcome “real”?
In a 2002 article, Moerman and Jonas
8 answer both questions in the affirmative. Presenting data from various studies, Moerman and Jonas
8 demonstrated that the subject-perceived meaning (or significance) of a situation can have an impact on outcomes for various somatic and psychological illnesses—as well as life expectancy. They propose relabeling these responses “meaning responses.” Their conclusion implies that study outcomes, even for pharmacologic agents or medical procedures, are affected by patient expectation. If this premise is true, the assumption of linear causality in today's gold standard study design is no longer unassailable.
In his still-pertinent 1991
JAOA special communication article, Irvin M. Korr, PhD,
9 wrote:
The philosophy of osteopathic medicine poses an indirect challenge to the linear cause-and-effect model presented by the popular RCT. By consistently and firmly placing the patient at the heart of the healing process, the osteopathic medical profession has long held that additional variables can have a dramatic effect on health outcomes. Some of these variables include patient perceptions and interpretations
3,5,6,8 as well as patient-physician interaction.
10 From our perspective, the double-blind RCT is self-limiting in its usefulness to researchers interested in investigating the efficacy of OMT. Although the RCT certainly can be useful in situations where a specific osteopathic technique is under investigation, it does not help researchers analyze all the questions the osteopathic medical profession wants to ask—and always has asked. In other words, a system cannot be analyzed by breaking it into its individual components and a priori assuming that some of those components are of no value or akin to epiphenomena. A system will almost always change its characteristics when taken apart.
However, the RCT does have its place in limited aspects of osteopathic medical research. This principle was amply demonstrated in a 1970s-era study conducted at the University of California, Irvine, College of Medicine.
11 In that study, subjects with acute low back pain received one of two manual interventions, either the high velocity/low amplitude treatment's lateral recumbent roll thrust or the control treatment, which consisted of the thrust lock only (ie, no actual thrust). Researchers reported an immediate improvement in subjects assigned to the active treatment group in several measures, such as straight leg raising, pain, and activities of daily living, such as sitting up in bed, reaching, and dressing. In the Irvine study,
11 a specific osteopathic manipulative
technique was being tested—not OMT as a system, or as an approach toward health maintenance. Studies of this kind are ongoing within the osteopathic medical profession and serve a useful function.
12,13
Thus, in OMT research, it is vital to know the kind of research question posed. If the question involves the effects of a single, isolable maneuver, it is possible to design a study roughly following the gold standard RCT model, with the notable exception being that the treating clinician cannot be masked to treatment, as observed by Licciardone and Russo
1 among other researchers.
4,14,15
However, if the research question involves the effect of OMT itself, it is simply not possible to factor out “placebos” because they are an integral part of the treatment, as noted by Korr
9:
Thus, to evaluate the system of OMT fully, either as to its effect or as to mechanisms, we must not attempt to pull out all but one single “active ingredient.” We must acknowledge that there are many such ingredients that comprise the total effect. Furthermore, in many cases, these ingredients may not be only interactive but multiplicative. To separate them may well destroy much of their effectiveness. “Therefore, that which is regarded as nuisance and source of error from one perspective is essence and source of clinical results from the osteopathic perspective,” concludes Korr.
9
And, indeed, there is still much controversy over the actual effectiveness of the double-blind RCT in demonstrating drug efficacy, as this model may routinely lead investigators to underestimate such effects.
3,5
However, osteopathic principles demand that research on the effectiveness and mechanisms of OMT recognize the centrality of the patient in creating the response called “healing.” Shared principles within the osteopathic medical profession similarly insist that healing is the result of the patient, not of a pharmacologic agent—or even OMT maneuvers.
16-18
I believe that osteopathic principles currently require us to rethink the meaning of the questions we are asking in regard to the effects of OMT as a therapeutic system. The patient must be the center of the process, and not relegated to the role of an epiphenomenon. Patients' expectations and experience interact with the skills of the physician to produce functional change. Thus, we must also search for better ways to measure function in our patients so that we can assess change, not just symptoms of a named disease. This process of asking the right questions of OMT will lead to better-understood comparison groups and a greater understanding of the actual effects of OMT on function—not merely on symptoms alone. Osteopathic manipulative treatment is an ongoing interaction between the unique structure and function of the patient and the skills of the physician, including the belief systems of both individuals. As such, it is dependent on these interactions for at least part of the treatment outcome. Analogously, listening to one instrument play its part in a symphony is in no way like listening to an entire orchestra play the same piece. Continuing the analogy, listening to an orchestra play a symphony under the direction of a novice conductor is entirely different from listening to the same orchestra led by a master.
With the intention of formulating ever more appropriate design models for osteopathic medical research, the osteopathic medical community must continue to rethink the questions it poses within the framework of evidence-based medicine as well as within the framework of the philosophy and practice of osteopathic medicine.