Dr. Christopher Moyer discusses the research behind massage therapy and cortisol levels at the Registered Massage Therapists’ Association of British Columbia’s Mental Health Symposium in at the Anvil Centre in New Westminster, British Columbia, April 2017. (Photo by Nick Ng)
Key takeaways:
- Prior plausibility: Evaluate how a claim aligns with established science. For example, “human energy fields” lack solid biological or physical foundations, making them low in plausibility—unlike massage’s impact on nervous and immune systems, which is far more plausible
- Methodology: Focus on the methods and results sections—they’re the least biased. Check the PICO framework: Population, Intervention, Control, Outcomes. Watch for participant drop‑outs; if many leave early, especially from the control group, it can skew results toward positive findings.
- Effect size: Don’t just look for statistical significance—ask how large the impact is in real‑world terms based on the Cohen’s d scale where: 0.2 = small; 0.5 = moderate; 0.8 = large.
Example: Meta‑analysis of massage on cortisol showed very small changes (d ≈ 0.12–0.15)
- Clinical equipoise: Researchers must enter trials without preconceived preferences or bias. True equipoise avoids unintentionally favoring one intervention over another—something difficult when practitioners know which treatment they’re delivering
- Translational relevance: Ensure findings from animal, cellular, or purely physiological studies actually apply to real‑world hands‑on massage settings
- Regression to the mean: The natural fluctuation of symptoms can exaggerate or undermine a treatment’ effectiveness. Individuals at extreme levels often regress toward average over time, regardless of intervention. Therefore, without proper controls and repeated measurements, improvements might be misattributed to the treatment.
Does massage therapy actually lower cortisol level in our body? If you search on Google Scholar or PubMed, you might find numerous studies that say yes. However, if you dig a little deeper and examine the literature on a broader scale, it actually says, “not really” or “nobody really knows.”
Psychologist Christopher Moyer broke down the conflicting research about massage therapy’s effect on cortisol levels at the Registered Massage Therapists’ Association of British Columbia’s Mental Health Symposium in 2017. On one hand, some research—primarily done at the Touch Research Institute (TRI) in Florida—found that a single dose or multiple doses of massage can reduce cortisol significantly.
But Dr. Moyer’s research team found “very little, if at all.”
Moyer et al. reviewed 19 qualified studies with a total of 704 participants (614 adults). The data points stretched across the chart, meaning that there is a high variability among the subjects.
Dots with long black horizontal lines that cross the vertical dotted line (the mean) tell us how strong or weak the effect of the treatments were.
Multiple-dose effect sizes and 95% confidence intervals.
Single-dose effect sizes and 95% confidence intervals.
In the graphs above, a long confidence interval means there is more uncertainty or less precision in the estimate. A shorter or narrower confidence interval means more precision and less uncertainty. Having a long confidence interval often means the trial has a small sample size, high variability in the data, biases, and/or varying effect sizes.
Massage research is still quite low in quantity and quality compared to physical therapy and nursing, and it still runs into many problems that other healthcare professions face, according to Moyer’s presentation. He suggested that massage therapists should not take the research they read for its face value and should critically appraise research within and without the field.
So what should therapists consider when they read massage research?
What should you fact check in massage research?
A 2019 editorial published in the British Journal of Sports Medicine (BJSM) suggested why physical therapists need to adopt a strong and honest evidence-based practice to avoid misleading the public and to draw attention to interventions that are effective and efficacious.
The authors gave an example where patients with acute low back pain should get physical therapy first as the first line of defense, according to the American Physical Therapy Association (APTA). However, this suggestion is not backed up by sound evidence.
In fact, the authors cited that such endorsements ignored the “high-quality evidence that early physical therapy is more costly and provides no benefit over usual care of [low back pain].” They also wrote that patients who get physical therapy are more likely to use opioids, imaging, surgery, and injections than those without physical therapy.
1. Prior plausibility
If your mom told you that a great white shark had attacked her boat while she went fishing, showed you the damage to the hull, and you know the waters where she fished is home to sharks, then you’d probably believe her.
But if she told you that a two-headed sea serpent had attacked her and showed you the same evidence, you’d probably ask for more evidence and fact check your mom.
This is basically what prior plausibility implies: How valid is a claim based on the existing body of knowledge? In that example, do two-headed sea serpents exist given what we know about marine biology and paleontology?
For example, the concept of a human energy field that is sometimes taught in massage therapy courses has a low prior plausibility because it lacks a well-established basis in current biology and physics. It doesn’t align with current scientific understanding of anatomy, physiology, and biophysics.
Since there’s no clear mechanism or empirical evidence showing that such an energy field exists or can be manipulated to produce consistent therapeutic effects, the prior plausibility for this claim is low.
However, the concept of touch having a therapeutic effect because of a combination of activities in the nervous and immune systems, skin receptors, and various confounding factors would be high prior plausibility.
2. Methodology: How was research done?
(Photo by Max Pexels)
Research methodology is a structured approach and set of procedures researchers use to collect, analyze, and interpret data to answer specific research questions or test hypotheses.
“The Introduction and Discussion sections are subjective and could be biased to make the study look favorable and important,” Dr. Anoop Balachandran said, who is the assistant professor of Exercise Science at Queens College at the City University of New York in New York City. “Remember that the methods and the results section are the only objective sections of a study.”
He suggested that clinicians follow the “PICO” of the study, which stands for population, intervention, control, and outcomes. Balachandran describes PICO as follows:
P for population
“It tells us what the population is used in the study. For example, it could be people with chronic pain, acute pain or hip or knee osteoarthritis, older adults, males and so forth,” Balachandran said. “Research is very population-specific, and an intervention that works for acute pain in older adults may not work well in younger or who have chronic pain.
I for intervention
“It is also important to understand the specifics of the intervention,” Balachandran said. “For example, if an exercise intervention, specifics—such as how many days/week, intensity or dosage, duration, and so forth—should be noted. The specifics of an intervention can make a big difference.
“For example, a 10-week intervention could have different outcomes compared to a similar intervention but with a longer duration of one year.”
C for control or comparator group
A control or comparator group informs researchers what they are comparing the intervention against.
“People always talk about the intervention but ignore the control group,” Balachandran said. “For example, a new exercise intervention could look spectacular if my control group is just a sedentary control, but it will look mediocre if I compare it to group that is given a standard exercise protocol.”
O stands for outcomes
“The primary outcome for manual therapists’ studies are pain or/and physical function. What people care about is if they feel better or move better,” Balachandran said. “For example, improving ROM (range of motion) is usually an outcome, but this outcome has no meaning in a person’s life whatsoever. This outcome only becomes meaningful if the improved [range of motion] helps with tasks that were previously impossible, like washing hair or wearing a jacket. So outcomes do matter.
“Once you have identified the PICO, you have managed to understand the bulk of the study in a very short time.”
Regardless of which profession the research is focused on, methodology applies to all randomized-controlled trials (RCTs), Balachandran added. Some of these factors are:
1. Did the participants have an equal chance of falling in either of the groups ( were they randomized)?
2. Could the participant or investigator have predicted which groups they will fall into (concealed allocation)?
3. Were the participants and investigator blind to the treatment (double-blind)?
4. Did they include all the participants in their final analysis (intention to treat)?
“In my opinion, the most important methodological concern for manual therapy studies is blinding,” Balachandran said. “Did the participants knew if they were getting the real or the sham treatment? Ideally, we want the sham treatment to look and feel exactly the same as the intervention minus the part that is supposedly giving the benefits of the intervention.
One example he gave is studies on acupuncture and its effect on certain pain relief. When such trials compare actual acupuncture versus fake acupuncture, both types show almost the same effects. (In fake acupuncture, the needle retracts into a sheathe after the needle contacts the skin.)
“Also, in good methodological studies, they also report if the blinding worked or not by asking the participants after the study,” Balachandran said. “Of course, the doctor is aware that he is giving the sham or the real treatment so there is a high risk of bias. So there is some bias that cannot be completely eliminated even in the most rigorous studies. In short, studies with good methodology should have a well-validated sham treatment and also report if the blinding worked based on the feedback from the participants.”
Balachandran also considered dropouts in a study that can affect its bias.
“This is one aspect that most folks ignore or are not aware of. It is now pretty clear that dropouts in a study could be related to the prognosis,” he said. “This simply means that people who drop out from a study might be the ones who didn’t experience any benefits or may have worsened their pain.
Basically, if people didn’t get better during the study, they would drop out. All that is left for the researchers to work with are those who responded well to the treatment.
“So the people who remain in the intervention group are the ones in which the treatment worked well, and naturally, the group averages tend to be better than it truly is,” Balachandran said. “This becomes more important when one group has greater dropouts than the other group.
“We will never know if the results of a study is the actual truth, but the better the methods, the more confident we are that we are closer to the truth. After all, science is all about finding the truth.”
3. Effect size: Does it tell us if a treatment is worth paying for?
The close the two comparisons, the less differences there are. Image by Skbkekas.
In research, an effect size tells us how much difference there are between two treatment groups, and knowing it can help recognize how effective and efficacious the treatment is.
If a research paper says that patients who were treated with Swedish massage had less low back pain than those who had received no intervention or had a different type of treatment, how big was the difference?
After all, researchers could claim a tiny difference between groups can be “better,” in which some people can market that intervention as “effective.”
Effect size, also known as Swedish massage, is measured on a scale where:
- 0.8 is ranked as “high”
- 0.5 as “moderate”
- 0.2 as “low”
Back to the Moyer’s study on the effect of massage on cortisol levels, the researchers pooled 11 studies on single-dose massage sessions as the first treatment in a series of treatments with a total of 460 participants and found that “[massage therapy] did not reduce cortisol significantly more than control treatments” with the effect size (d) of 0.15.
When they examined eight studies of single-dose sessions with 307 participants as the last treatment of a series, they also found the same result (d = 0.15).
Among multiple-dose massage sessions, they also found low effect sizes (d = 0.12) among 16 studies with a total of 598 participants.
When these studies are split into adults and children, the adults (n = 508) effect size dropped more than 50 percent (d = 0.05).
However, the children group ranked higher with a “moderate” d = 0.052. The last result may seem promising, but Moyer et al. cautioned that this data is based on a tiny number of studies and likely is “vulnerable to the file-drawer threat” where many studies with negative outcomes are not published.
Also, larger effect size does not always mean a treatment is “better,” nor does a smaller effect size mean it is “not very good.” Context is needed to make sense of the data’s meaning.
For example, if being active regularly can improve mortality by 0.09 over five to six years, this small effect can still save many lives by encouraging people to be more active.
But if a treatment only reduces back pain by 0.09 when compared to standard care or placebo, then it may not be in the best interest for patients to get and pay for that treatment.
4. Clinical equipoise: How much bias are there?
In some ways, manual therapy research is similar to a Mortal Kombat game where one modality is “pitted” against another to see which is “better.” This bias not only may lead researchers the everything-is-a-nail mentality, it also can skew the outcomes of manual therapy research.
This is where clinical equipoise is needed, which is the assumption that one intervention is not better than another.
Physiotherapists Chad Cook and Charles Sheets wrote that to have true equipoise, the researchers must have “no preference or is truly uncertain about the overall benefit or harm offered by the treatment to his/her patient. In other words, the clinician has no personal preconceived preferences toward the ability of one or more of the interventions to have a better outcome than another.”
They wrote with the best intentions, researchers may unconsciously set up the experiment in a way that favors the tested intervention or undermines the control group.
It’s also possible that the “placement of importance, enthusiasm, or confidence associated with one’s expertise in an intervention” can influence the experiment’s outcome, which also includes how well the patients perceive the treatment.
Having pure personal and clinical equipoise may be unlikely to avoid to do because it’s impossible to blind the practitioner. It’d be like working blind-folded while providing the treatment, which isn’t practical.
5. Can the massage research be applied to hands-on work?
Applying therapeutic exercise research to practice at the 2017 San Diego Pain Summit with Ben Cormack. (Photo by Nick Ng)
In the late 1970s to early 1980s, biologist Christiane Nüsslein-Vollhard made some discoveries about how fruit flies grow and develop with her colleagues Dr. Eric Wieschaus and Dr. Edward B. Lewis. She had no idea that their research on fruit flies could end up being one of the main staple animals that many biomedical research use in the next three decades.
What they did was basic science — research done for the sake of understanding a specific phenomenon with no or little consideration for its real-life applications.
What came after was translational science, using fruit flies for a variety of scientific disciplines, including evolutionary biology and behaviorial genetics that could identify certain genes that contribute to human diseases.
But in manual therapy research, Dr. Neil O’Connell—a physiotherapy researcher at Brunel University—said that clinicians should should be careful about extrapolating basic research to hands-on practice since the researchers may not often have practicality of their work in mind.
Conversely, sometimes translational research in manual therapy don’t consider basic science and may come up with false positives, he added.
“There is a tendency to not scrutinize basic mechanisms of research through the same critical lens that we do clinical trials,” O’Connell said. “But this is a mistake as the field is troubled by small sample sizes and avoidable biases that are not controlled.”
O’Connell cited a 2011 acupuncture study that finds sticking needles in certain parts of the forearm can change how the brain perceives pain. However, such changes does not mean the interventions work.
“Brain activation in response to needling neither validates acupuncture nor provides a cogent mechanism for therapeutic action,” he said. “That it activates sub-cortical networks involved ‘endogenous pain modulation’ also does not distinguish acupuncture as a possibly effective active treatment since these networks are also implicated in the placebo response.
“It is not that the data in this study tells us nothing important. It adds to an existing body of data regarding how sensory input is cortically processed. But as a study of therapeutic mechanisms, it attributes possible mechanisms to an effect that the best evidence tells us does not exist.”
O’Connell said that somehow finding a physiological effect of a treatment legitimize the intervention, and many therapists and companies would likely market this to the public because research found “something” that seems to “work.”
“We tend to forget that even when a biologically plausible mechanism exists, it is not a given that it will translate into a meaningful clinical effect,” he said. “Think of the scale of failed new pharmaceuticals where the biological plausibility work all looked good. None of this is to denigrate the value of various types of evidence. The danger lies in taking any evidence, not matter how removed from the patient, as some form of tacit legitimization of your therapy du jour.
Researchers might take a finding in a cell or animal model and use it as evidence to support a treatment’s effectiveness, O’Connell cautioned, adding that the effect may likely in indirect.
“They will often presume that any evidence of some physiological change is evidence that a treatment is plausible,” O’Connell said.
6. Do positive outcomes validate a treatment?
Tying back to outcomes, when we see our patients feel better after a treatment, we often give credit to our treatment for making them feel better. But this alone cannot be a validation to the positive (or sometimes negative) outcomes because outcomes only measures outcomes, not the effects of the intervention.
This is because many factors other than the intervention can affect the outcome, such as the non-verbal interactions between the therapist and the client or patient, the natural regression of pain, and patients’ beliefs and prior experience to the intervention and interaction with their previous therapist.
Sometimes a positive outcome can happen without any interventions; and we can also say the same thing about negative outcomes which doesn’t always mean that the intervention “sucks.” It could have gotten worse without the intervention.
Relying on outcomes can also blind us from acknowledging those who do not respond well to the treatment, which is a demonstration of survivorship bias—people tend to notice and acknowledge successful outcomes while sweeping aside unsuccessful ones.
Even though high quality randomized-controlled trials are the best method to use, Herbet et al. suggested that therapists should avoid the being extreme on either side of the decision-making process, where one side relies on outcomes and the other side relies strictly on randomized controlled trials. We need both.
When there is no good available evidence, clinical outcomes are likely all we get to use. That could be risky, depending on the nature of the problem and the type of intervention.
However, when randomized-controlled trials “provide clear evidence of the effects of an intervention from high quality clinical trials, clinical outcome measures become relatively unimportant and measures of the process of care become more useful,” Herbet et al. wrote. “When evidence of effects of interventions is strong, we should audit the process of care to see if it is consistent with what the evidence suggests is good practice. When there is little or no evidence, we should audit clinical outcomes.”
Related reading: Why reading scientific abstracts of research paper could mislead you
“I recently heard a man interviewed on my local public radio station complain about the difficulty of keeping up with what he called the ‘swerves of scientific wisdom’: ‘I spent two tours in Iraq as a gunner,’ he said, ‘and I know how hard it is to hit a moving target. I wish these scientific experts would just hold still.’
“But that’s the thing. Holding still is exactly what science won’t do.” ~ David P. Barash, evolution biologist and psychology professor, Aeon
Nick Ng is the editor of Massage & Fitness Magazine and the managing editor for My Neighborhood News Network.
An alumni from San Diego State University with a bachelor’s in graphic communications, Nick also completed his massage therapy training at International Professional School of Bodywork in San Diego in 2014. In 2021, he earned an associate degree in journalism at Palomar College.
When he gets a chance, he enjoys weightlifting at the gym, salsa dancing, and exploring new areas in the Puget Sound area in Washington state.
