J Manipulative Physiol Ther 1999 (Sep); 22 (7): 436–443
Cheryl Hawk, DC, PhD, Ayla Azad, DC, Chutima Phongphua, MD, DC, Cynthia R. Long, PhD
Palmer Center for Chiropractic Research, Davenport, Iowa 52803, USA
OBJECTIVE: To identify aspects of the delivery of placebo chiropractic treatments by using sham adjustments that may cause a treatment effect and that may affect the success of blinding.
DESIGN AND SETTING: Two-period crossover design in a chiropractic college research clinic.
SUBJECTS: Eighteen volunteer staff, students, and faculty of the chiropractic college who reported low-back pain within the last 6 months.
INTERVENTIONS: Flexion-distraction technique was used to perform chiropractic adjustments, and a hand-held instrument (Activator adjusting instrument) with the pressure gauge set on the 0 was used to perform sham adjustments. The treatment period was 2 weeks, with a total of 4 visits.
MAIN OUTCOME MEASURES: The Visual Analog Scale (VAS) for pain and Global Well-Being Scale (GWBS).
RESULTS: Although VAS and GWBS scores improved with both treatments, a somewhat greater improvement occurred in most cases with the active treatment. Eight of 14 patients interviewed believed that the placebo had a treatment effect.
CONCLUSION: This study provided preliminary information that was useful in planning the protocol for a placebo chiropractic treatment in the randomized clinical trial for which it was designed.
From the FULL TEXT Article
This study was designed to provide preliminary information to be used in planning a placebo chiropractic treatment in a randomized clinical trial. It had several limitations that must be considered when the results are viewed. First, it was specifically designed to test only the procedures planned for our clinical trial, so its findings should only be applied to studies of similar design by using similar procedures. Second, its sample size was small, so significance testing was precluded as a result of limited power. For this reason, we intentionally did not aggregate the data or perform a statistical analysis. Furthermore, the randomization of patients in this study into 2 time-sequence protocols (active treatment followed by placebo or placebo treatment followed by active) resulted in groups that were not equivalent in their baseline VAS scores for low back pain, further reinforcing our decision not to aggregate the data. Third, since this was a crossover design, there may have been carryover effects from the treatments.
In regard to the specific aims of the study, information was obtained related to patients' perceptions of and symptomatic response to the active and placebo treatments and to what impact the sham-adjustment portion of the placebo treatment had on patient perceptions and responses. We included both symptomatic and asymptomatic patients as well as patients who were aware of placebo treatment. Four of the patients who knew which treatment was the placebo were so informed to increase the influence of the information on patients' perceptions of treatment effects.
The information most important for use in planning a clinical trial is that 8 of 14 patients unaware of this issue believed that the placebo treatment had an effect. None of the 3 patients who knew that one of the treatments was a placebo correctly identified which it was. Two of these patients (28 and 29) thought the active treatment was the placebo, even though both their VAS and GWBS scores improved more with the active treatment. Their inability to discern the placebo supports the idea that the active treatment effect is not the result only of patient belief in it. These findings indicate that the placebo treatment we used may have some degree of success in blinding patients to group status.
Furthermore, patients provided helpful information on which aspects of the placebo treatment made them feel the treatment was a real one and which aspects of it may actually have been a real one. Simply having a standard of comparison may be one factor, which we inferred because nearly all the patients who thought placebo treatment had no effect had received the active treatment first. However, an alternative interpretation for this finding is that the patients who received the active treatment first and made the greatest improvements (Table 1) received the placebo after they already felt much better. Therefore, significant improvement was less likely with the placebo than with the active treatment. Most patients believed that the sound made by the adjusting instrument was helpful in conveying the impression that active treatment had occurred. The palpation aspect of the placebo treatment produced muscle relaxation, a non-specific treatment effect to be expected as a result of any hands-on treatment. Similarly, the doctor's attitude and interaction with the patients may be expected to contribute to the nonspecific placebo effect.  In our upcoming study, both of these effects will be standardized across groups by standardizing the amount of time doctors spend with patients in all groups and the amount of pressure used in palpation or light massage. Two patients believed that the sham adjustment may have actually produced a specific correction to the vertebrae; it is important to guard against this effect even in a minority of patients, so as to avoid a misclassification error in the future trial. Consequently, we will address this issue by prescribing the exact parameters for delivery of the sham adjustment: 1) the sham adjustment should be delivered on the paraspinal musculature, not over the portion of the vertebra used in delivery of a real adjustment; and 2) a predetermined amount of pressure should be applied with the instrument, because it appeared that even at a 0 pressure setting the doctor's manual pressure may have been enough to cause a correction in certain patients.
It is important to note that the placebo treatment had a treatment effect. This is consistent with the findings about placebo treatments in studies of other hands-on and pharmaceutical therapies as well. [1, 2, 22] However, the patients in our study indicated in their interviews before unblinding that the placebo effect was shorter lasting and not as great. The VAS and GWBS findings support the less substantial treatment effect of placebo; however, because patients in the 2 time-sequence protocols were not equivalent in terms of baseline low-back-pain intensity, we could not gather much information on the comparative persistence of treatment effects. It is important to consider the possibility of placebo effects being fewer and of shorter duration, which would indicate the importance of considering the anticipated benefit from the specific effects of the active treatment above that of the non-specific effects of the placebo treatment when calculating an adequate sample size. This also indicates the importance of long-term follow-up.
These results indicate that we need to further refine our placebo protocol to minimize specific effects while still maintaining blinding by using the procedures discussed above. This will entail conducting a training period in which the clinicians involved in the trial will practice delivering measured forces during the adjustment, the sham adjustment, palpation, and light massage by using a force plate to measure force and by using qualitative patient feedback on the effects of various levels of force. This training is currently underway as part of the multisite clinical trial pilot.