Our findings suggest that many patients would be willing to try specific CAM therapies for back pain, especially if they had high expectations for their helpfulness. Interestingly, we found no consistent relationships between high expectations for a particular therapy and either previous use of that therapy or high self-perceived knowledge of that therapy.
Our findings regarding knowledge, previous use, and expectations for these therapies were largely similar for Seattle and Boston and for older and younger adults. However, those over 65 years old were less likely to have high expectations of acupuncture and massage and to have tried massage previously.
Since we conducted the study in two metropolitan areas where CAM use is fairly common, our results might not represent the CAM views of patients with back pain in more rural areas or in other regions of the country. Another limitation of our study was that 30% of people we attempted to contact could not be assessed for eligibility, leading to the possibility of a high non-response rate. Because we have almost no information on the characteristics of the individuals with unknown eligibility, we do not know if they differ from those included in the study, and cannot adequately estimate the magnitude and direction of potential biases regarding interest in CAM that might exist in our sample. However, the fractions of individuals who were unable to be assessed for eligibility were similar among those less than 65 years of age and those 65 and older in each metropolitan area (45% vs. 40% in Boston, respectively; 20% vs. 20% in Seattle).
Respondents showed a clear preference for receiving hands-on treatments delivered by a practitioner compared to attending classes that teach self-care techniques. Whether this reflects a preference for provider-oriented, more passive, therapies or the belief that classes teaching these specific self-care therapies would be less effective is not clear. Unfortunately, our interview did not include questions about yoga, which has recently received more popular press than meditation or t'ai chi as a self-care therapy for back pain [9,10].
Survey respondents were not enthusiastic about "meditation training" as a treatment for back pain. Relatively few of those who indicated prior use of meditation for physical or mental health problems had used the forms of meditation most commonly taught in a medical setting (e.g., mindfulness meditation). Consequently, studies recruiting patients to participate in interventions including meditation training may need to carefully describe the treatment in terms of a concrete goal (e.g., stress reduction).
There is still relatively little knowledge about and experience with acupuncture and t'ai chi even in Boston and Seattle where use of CAM therapies is generally high. In fact, about one – quarter of respondents were unable to provide an expectation of the helpfulness of acupuncture or t'ai chi. Nevertheless, substantial fractions of participants were willing to try acupuncture and t'ai chi as a treatment if their primary care provider thought it reasonable, and in the case of acupuncture, even if they had to pay a $10 co-pay each visit. Our finding that people in our sample reported being almost as willing to try acupuncture as massage, despite less knowledge of, expectations about and experience with it, is intriguing and requires further inquiry.
Although participants in this study reported more knowledge of and experience with chiropractic, they were more enthusiastic about massage. A recent survey of 46,000 Consumer Reports subscribers found that among those who had experienced back pain, the relatively few who had tried deep tissue massage rated it more favorably than those who had tried medications or physical therapy. The use of massage in this country has been growing steadily since the 1960's, with the largest increases in the 1990's . In fact, in surveys of CAM use in the US population conducted in 1990 and 1997, Eisenberg et al.  found that massage as a treatment for various medical conditions had increased 61% over the seven-year period, while chiropractic remained fairly stable. By 1997, the estimated percentage of US adults who had used chiropractic was similar to that who had used massage, 11%. The relative popularity of massage may result from the more positive experiences of those who have tried it compared with chiropractic or acupuncture, and higher expectations that massage would be helpful for their current pain. Moreover, chiropractic users were more likely to report treatment related "harm" or "pain" than were users of massage.
Implications for clinical trials
Most survey respondents indicated they were "very willing" to participate in our two hypothetical clinical trials evaluating different CAM treatments for chronic back pain. Massage was the preferred treatment in both trials, but more than one in five survey respondents stated a preference for acupuncture and t'ai chi. In view of the long-standing popularity of chiropractic, surprisingly few respondents reported chiropractic as their top choice. Nonetheless the finding that massage was substantially more popular than chiropractic mirrors the results among acute low back pain patients in a clinical trial who were randomized to a choice of acupuncture, chiropractic, massage, or usual care or to usual care alone . In that study, 52% of the participants said they would choose massage if given a choice, compared with only 24% who said they would choose chiropractic if given a choice. This finding could reflect the fact that many people have access to chiropractic as part of their current health care coverage .
Despite low levels of knowledge about t'ai chi and acupuncture, the finding that over 40% of respondents indicated they were very likely to try these therapies suggests that recruiting enough subjects for clinical trials involving these therapies may be feasible if moderate to large patient populations are available. Recruiting patients for meditation trials, however, is likely to be difficult. Consequently, when we recruited patients for a pilot trial that included a stress reduction intervention based on the principles of mindfulness meditation, we chose to describe it as "Mindfulness Based Stress Reduction" rather than mindfulness meditation.
We believe that clinical trials evaluating obviously different treatments for chronic low back pain, such as massage and meditation, may have problems retaining subjects who do not receive the treatment (e.g., massage) that attracted them to the study. This problem may be exacerbated if patients have an exceptionally strong preference (or dislike) for one treatment. Inclusion of multiple CAM modalities in a single study risks tempting potential participants to sign up for the study in the hope of receiving a desired treatment, and then dropping out if they receive a different treatment.
In addition, if one treatment is vastly more popular than another, it could be difficult to disentangle the effects of patient expectations and treatment efficacy per se, leading to difficulties in interpreting positive study outcomes. This problem is compounded by concerns about the subjective nature of back pain outcomes, the difficulty in masking participants to study treatment, and the strong skepticism of some researchers that CAM treatments can be effective, even when results are impressive. Masking patients to treatment is quite difficult in studies of many types of conventional as well as CAM treatments if the treatments involve a physical modality, such as massage, or active participation of the patient in the treatment, as in t'ai chi. In such circumstances, using masked outcomes assessors is important to minimize bias. We also suggest that patient (and provider) expectations for treatment and prior experience with each treatment, be measured and, if appropriate, controlled for in the analyses. Finally, if a particular therapy is shown effective in clinical trials in different populations, mechanistic studies will be important for determining how these therapies achieve their effects. Such studies are especially important to convince skeptics that CAM therapies actually have specific effects. In the meantime, the high and rising public interest in CAM therapies, especially for musculoskeletal conditions , highlights the importance of evaluating the effectiveness of various CAM treatments for back pain and our findings suggest that recruiting for these efforts may not be difficult.