Under the Microscope

A Meta-Analysis of Massage Therapy Studies

By Shirley Vanderbilt

Originally published in Massage & Bodywork magazine, June/July 2004.

Practitioners and clients espouse the strengths of massage therapy. But what happens when massage studies are subjected to a rigorous meta-analysis? Are the benefits confirmed?

A University of Illinois research team decided to find out. The fact that they were psychologists deserves a notation. It seems massage may offer benefits similar to psychotherapy when it comes to decreasing anxiety and depression, an exciting find says Christopher Moyer, whose research recently appeared in Psychological Bulletin (January 2004). But before we begin to analyze the study, let’s make it clear Moyer’s findings do not suggest the substitution of massage, as a stand-alone modality, in place of professional psychological or medical treatment for these conditions. “We may find it is a good complement for (treatment of) depression,” he says, especially in combination with other forms of care.

As a competitive cyclist, Moyer appreciated the value of massage in training, and this, combined with his work as a graduate student in psychology, sparked an interest in examining the effects of massage therapy. A literature review revealed that of those meta-analyses on record, none had “quantitatively reviewed the range of commonly reported MT [massage therapy] effects in physically mature individuals.”1

The meta-analysis by Moyer et al. is the broadest in scope of its kind. “We were hoping to fill a niche in the research,” he says, “and see if some of the claims made in individual studies would hold up. In general, they did, with a few exceptions.” That such large effects on depression and trait anxiety were substantiated “gives us a whole new way to look at and think about the treatment.”

In addition to evaluating benefits commonly reported in massage research, the team evaluated the impact of specific moderators, such as therapist training and minutes of therapy application. As a psychologist, Moyer was especially interested in commonalities between massage and psychotherapy. In order to make this comparison, the team eliminated studies using subjects too young for verbal interchange and/or understanding of an empathic relationship. Three other criteria were set for inclusion: use of between-groups design (meaning comparison of massage with one or more non-massage control groups); randomization; and sufficient data to generate a between-groups effect on at least one dependent variable.2

Utilizing the definition of massage therapy as “the manual manipulation of soft tissue intended to promote health and well-being,”3 the team identified 144 studies through a literature review. Only 37 met the inclusion criteria. These represented a total of 1,802 subjects, of which 795 were massage recipients. Comparison groups were categorized into wait-list subjects to include standard care, resting, reading, or work break (49 percent), and those receiving another active therapy or placebo treatment (51 percent).4

Studying the Studies

With the premise that massage therapy effects are both physiological and psychological in nature, the team makes a distinction between single-dose and multiple-dose effects. Although pointing out these terms are not commonly used in research, they note a similarity to the “short-term” and “long-term” designation from the Touch Research Institute (TRI) of Miami, Fla.5

Single-dose effects are those that affect transient states: state anxiety, negative mood, pain, cortisol levels, blood pressure, and heart rate. Multiple-dose effects involve those “variables that are typically considered to be more enduring,” or influenced by a more prolonged treatment course as opposed to single application. These include trait anxiety, depression, and delayed assessment of pain.6

Of the nine variables (effects) investigated, the strongest findings were for depression and trait anxiety, leading researchers to state, “Considered together, these results indicate that MT may have an effect similar to that of psychotherapy.” In terms of statistical numbers, “The average MT participant experienced a reduction of trait anxiety that was greater than 77 percent of comparison group participants,” and for depression the reduction was greater than 73 percent of comparison group subjects, closely resembling meta-analysis findings for absolute efficacy of psychotherapy.7

Three of the single-dose effects reached statistical significance. For state anxiety, reduction was greater than 64 percent of participants in comparison groups, with similar findings for blood pressure and heart rate (respectively, reductions greater than 60 percent and 66 percent of subjects in comparison groups). Although a number of individual TRI studies have documented cortisol decrease, this effect did not reach significance.8 According to Moyer, this is not to say there was no effect, but within this particular analysis, the size of the effect in comparison with other groups was not remarkable.

In contrast to the lack of significant finding for effect on immediate assessment of pain, delayed assessment of pain was a significant finding. For subjects assessed several days or weeks after their course of massage treatment, levels of pain “were lower, on average, than 62 percent of comparison group participants.”9

Of the six potentially moderating variables (minutes of therapy, mean age, gender, type of comparison group, therapist training, and laboratory effect), none were determined to be significant. These results were not surprising to the team, with the exception of minutes of therapy and comparison group. The team had sought to “determine whether there are minimum or optimum dosages of MT,” and had also anticipated finding a greater effect in comparison to no treatment than to active/placebo groups.10

While the findings did not support a relationship between minutes of therapy and effect, the authors note this may be due to the insufficient statistical power of the studies within hand, as it would seem logical that longer doses of an effective treatment would produce an even more potent effect. “Nevertheless,” the team writes, “it must be concluded that this moderator may not be as important as we predicted, and that even short sessions of MT can be effective.” Although no studies to date have examined effects of session length, incorporating two levels of this variable in future research could “more powerfully” determine its influence.11

Within the studies measuring anxiety and depression, none incorporated a design of combining wait-list and active/placebo groups within the same study. The influence of comparison groups may be made more evident by incorporating this type of design into research on anxiety and depression,12 but Moyer places more importance on examining how massage brings about these effects.

In the studies cited, qualifications of those applying massage were not specified beyond identification as trained practitioner (65 percent), minimally trained person (22 percent), and no indication of level of training (14 percent). The negative finding for moderating influence, authors write, should not be taken to mean training is inconsequential, as it was not feasible with the information provided to determine the level of expertise. But they do point out that these positive findings indicate even “laypersons provided with some training can provide beneficial MT, information that may be valuable to researchers working with limited resources.”13

TRI has been a leader in massage research, producing a large proportion of the studies in existence to date. But despite the fact that TRI studies constituted 32 percent of those analyzed, there was no evidence of this single laboratory as a moderating influence.14

Puzzles and Pieces of Data

Identifying the explanatory mechanisms behind the effects of massage, the authors write, has received little emphasis in previous massage research. In this review, they attempt to reconcile their findings with those theories predominant in the literature. Gate control theory of pain reduction, the most frequently cited theory, was least supported by the analysis, as substantiated by the results on immediate assessment of pain. This theory holds that massage produces an analgesic effect by supplying sufficient pressure to block the gate, so to speak, before painful stimuli can be experienced and processed by the brain.15

In contrast, the authors state the positive findings for delayed assessment of pain may be more related to the theory that massage reduces pain through facilitation of restorative sleep, although “without data on sleep patterns, this possibility is only conjecture.” Previous research has suggested a connection between sleep deprivation and increases in pain.16

Massage is thought, by some, to bring about a shift in the autonomic nervous system from a sympathetic response, or heightened sense of alert, to a parasympathetic response, or state of calm and relaxation. The theory of promotion of parasympathetic activity is bolstered, authors write, by the results on decreased blood pressure and heart rate. But the fact that no significant effect was found for cortisol reduction (an expected presence in a parasympathetic response) adds inconsistency to results. As for theories of mechanical effects and influence on body chemistry, neither was supported by this particular meta-analysis.17

While these mixed results leave us with no clear-cut answers regarding the more popular theories, the team holds open the door, suggesting, “MT’s effect on state anxiety, trait anxiety, and depression may come about as a result of MT’s influence on body chemistry, whereas the ability of a course of MT treatment to provide lasting pain relief may result from the mechanical promotion of circulation and breakdown of adhesions, or from improved sleep promoted by the treatment.”18

It is important to keep in mind that the intent of a meta-analysis is to summarize, or as Moyer says, “To see where we are so far.” That one particular theory is not substantiated by these results does not mean it has been disproved, he notes. For example, further research on the gate theory, with a specified condition and treatment, may indeed provide a supportive body of work for this explanation. By addressing these issues within the context of the available data, the meta-analysis not only allows us to see where we are thus far, but also helps us plot a future path.

Common-Factors Model

We now turn to the issue of interpersonal contact or attention. While this receives some interest in studies, the authors write, it is more likely to be considered “a nuisance variable, and comparison treatments are selected in such a way that different groups receive the same amount of attention.”19 But what if interpersonal attention itself is part and parcel of the effect?

Another way to look at the effects of massage, the authors propose, is through the perspective of psychotherapy’s common-factors model. In this model, more importance is given to factors such as client expectations, positive regard and warmth on the part of the therapist, and the alliance developed between client and therapist, rather than the specific modality applied. Likewise, benefits from massage may be related to similar elements, including “therapist’s personal characteristics and expectations.”20

Pointing to the size of effects on anxiety and depression, lack of influence of training level as a moderator, and commonalities between psychotherapy and massage in the structure of therapy sessions, the team writes, “Several of the findings in the present study are consistent with such a model applied to MT.” This suggests the similarities between psychotherapy and massage deserve consideration, especially in future research design. “Different questions need to be asked, different moderators tested, and different comparisons made.”21

The team highlights a host of issues to be addressed, especially regarding the process and intricacies of interpersonal contact, should massage research take this route. Comparison groups are another conundrum. “When a common-factors model is applied to MT,” the authors write, “the notion that a comparison treatment such as progressive muscle relaxation controls for attention is incorrect.” While there is no difference in quantity of attention, the quality of attention is not identical.22

Lest we be left with the impression that massage effects might be considered as purely psychological, the team clarifies by stating, “Clearly MT is at least partially a physical therapy, and some of its benefits almost certainly occur through physiological mechanisms. In fact, one of the most interesting aspects of MT is that it may deliver benefit in multiple ways; specific ingredients and common factors may each play a role, with each being differentially important depending on the desired effect.”23 And herein lies a crucial element of massage therapy research — not just what is happening, but how and why.