A four-tier framework for matching the strength of a wellness claim to the evidence behind it
Practices built on light, field, and current occupy an unusual evidentiary position. Some of their underlying biology is standard physiology. Some of their clinical applications have been tested in randomized trials with conflicting results. Some rest on models that are plausible but untested in the setting where they are offered. And some produce experiences that people value without any clinical claim attached at all. A single phrase such as “evidence-based” cannot describe all four states at once, and the recurring temptation in the wellness field is to borrow the credibility of the first state for claims that belong to the third or fourth.
The Evidence Gradient is the framework Mountain’s Daughter Health uses to keep those states apart. I introduced it briefly in One Body, Three Signals. This piece sets it out in full for readers who work with evidence professionally: where it comes from, how each tier is defined, how claims are assigned, where it falls short, and how it could be tested.
A note on the kind of contribution this is. The Evidence Gradient is a conceptual synthesis applied to practice: a vocabulary for labeling public claims, assembled from existing work on evidence appraisal, causal inference, and regulatory language. It is not a new empirical finding, and it has not been validated as a measurement instrument. Those limits are discussed below, and they shape how the framework should be read.
What the Gradient borrows
Evidence hierarchies are familiar to anyone trained in health research. The GRADE approach, now used by guideline developers worldwide, rates the quality of a body of evidence from high to very low and, crucially, keeps that rating separate from the strength of any recommendation drawn from it (Guyatt et al., 2008). Murad et al. (2016) later reframed systematic reviews and meta-analyses as tools for consuming evidence, that is, for appraising, synthesizing, and applying it, rather than as the fixed summit of a pyramid of study designs.
Two features of that literature matter here. The first is GRADE’s candor about its own categories: “Quality of evidence is a continuum; any discrete categorisation involves some degree of arbitrariness” (Guyatt et al., 2008, “How Should Guideline Developers Alert Clinicians” section). The authors accept that arbitrariness in exchange for simplicity, transparency, and vividness. The Evidence Gradient makes the same trade, knowingly.
The second is GRADE’s treatment of indirect evidence. Guyatt et al. (2008) recount how antiarrhythmic drugs were licensed because they suppressed asymptomatic arrhythmias, a surrogate that “reflected only indirectly on the outcome of sudden death” (Guyatt et al., 2008, “What Is Quality of Evidence” section); a later randomized trial showed that the drugs increased the risk of sudden death. Indirectness bears directly on wellness practice, where most available trials study clinical populations (osteoarthritis, fractures, chronic wounds) while the practice itself serves people seeking rest and restoration.
The same example anchors Howick et al.’s (2010) argument about mechanistic reasoning. They distinguish mechanisms, the arrangements of parts that are thought to produce stable relationships between inputs and outputs, from mechanistic reasoning, the inference from those mechanisms to a patient-relevant outcome. As they put it, “described mechanisms do not amount to evidence” (Howick et al., 2010, “Mechanisms and Mechanistic Reasoning” section). In their account, mechanistic reasoning can count as evidence of efficacy only when its inferential chain is complete and it takes complexity into account. Russo and Williamson (2007) reach a related conclusion from the philosophy of science: the health sciences make causal claims on the basis of evidence of mechanisms and evidence of probabilistic dependence together, so an account of causality built on only one of the two misdescribes how those sciences actually reason.
Regulation supplies a third boundary. The U.S. Food and Drug Administration’s (2026) guidance on general wellness products separates intended uses that support functions associated with a general state of health without reference to disease from those that make such a reference, and it states plainly that falling under the policy is not a finding that a product is safe or effective for its intended use. A wellness framing, in other words, is a statement about intended use. It is not a statement about evidence.
Finally, the arts-and-health literature shows how varied the evidence behind experiential practice can be. Fancourt and Finn’s (2019) scoping review for the World Health Organization drew on over 900 publications, including more than 200 reviews covering over 3,000 studies, and described designs ranging from uncontrolled pilots and single case studies to longitudinal cohorts and randomized controlled trials, with acknowledged variation in quality.
The Evidence Gradient takes something from each of these: GRADE’s separation of evidence from recommendation and its honesty about categories, Howick and colleagues’ caution about mechanisms, Russo and Williamson’s insistence on two kinds of evidence, the regulator’s line between wellness and disease, and the arts literature’s respect for experience as an object of study in its own right. What it attempts to add is a single public vocabulary that places mechanism, clinical outcome, theory, and experience on one page, with a form of language permitted at each level. Whether that adds something the existing frameworks do not already provide is a fair question for readers who know them well.
The four tiers
Each tier below has an entry criterion, a form of language it permits, and a boundary it does not license. The tiers classify claims. They do not classify modalities, devices, or practitioners, and a single device can carry claims at several tiers at once.
Tier I: Established mechanisms
Entry. The claim concerns a biochemical or electrophysiological process that has been replicated across laboratories and appears in standard physiology. Nerve conduction, the resting membrane potential, and the absorption of red and near-infrared light by mitochondrial cytochrome c oxidase belong here (de Freitas & Hamblin, 2016).
Permitted language. Descriptions of how a process works: “cells maintain an electrical charge across their membranes”; “a mitochondrial enzyme absorbs these wavelengths.”
Boundary. A tier I claim never licenses an outcome claim. This is where wellness communication most often fails, and photobiomodulation shows the gap precisely. That cytochrome c oxidase absorbs light in this range is well established. The account of what follows, that photons dissociate inhibitory nitric oxide from the enzyme and increase electron transport and ATP production, is described by de Freitas and Hamblin (2016) as “the leading hypothesis,” in a review that opens by noting that the field “still has not gained widespread acceptance, largely due to uncertainty about the molecular, cellular, and tissular mechanisms of action” (Abstract). The claim “a mitochondrial enzyme absorbs red light” is tier I. The claim “red light restores your energy” is not, and borrowing the first to sell the second is the error Howick et al. (2010) describe. Mechanism is where honesty starts, not where it ends.
Tier II: Mixed evidence
Entry. Comparative clinical studies exist, including randomized trials or meta-analyses, but their results diverge across outcomes, comparators, populations, follow-up periods, or device parameters.
Permitted language. Specific findings attributed to specific studies, with the disagreement stated: “some trials in osteoarthritis report improved function; results for pain are inconsistent.”
Boundary. A tier II claim cannot be stated as settled, and it cannot be carried from the clinical population studied to a wellness population without saying so.
Pulsed electromagnetic field (PEMF) therapy for knee osteoarthritis shows what “mixed” means in practice. Chen et al. (2019) pooled eight placebo-controlled randomized trials (421 patients) and found a statistically significant improvement in physical function (WMD = −5.28, 95% CI [−9.45, −1.11], p = .01) but no statistically significant advantage for WOMAC pain (WMD = −1.06, 95% CI [−2.30, 0.17], p = .09) or stiffness (WMD = −0.50, 95% CI [−1.09, 0.09], p = .10). Yang et al. (2020), pooling 15 placebo-controlled trials across osteoarthritis sites, reported benefit for pain (SMD = 1.06, 95% CI [0.61, 1.51]), stiffness (SMD = 0.37, 95% CI [0.07, 0.67]), and function (SMD = 0.46, 95% CI [0.14, 0.78]), and found that PEMF parameters did not influence symptoms. Viganò et al. (2021), across 13 studies and 914 patients, found a short-term reduction in pain (SMD = −0.41, p = .009) but no statistically significant overall change in activity (SMD = −0.45, p = .09). In their analysis, effects were larger against placebo, disappeared when PEMF was compared with alternative treatments, and shrank with longer follow-up, and the authors concluded that PEMF is not superior to conservative therapies such as physiotherapy. (Sign conventions differ between these reviews; the direction of each estimate favors PEMF where benefit is reported.)
Two cautions follow. First, statistical significance in these reviews is not practical significance. A pooled difference in a WOMAC function score can clear p < .05 without telling a reader whether the change is one a patient would notice. Second, these reviews do not simply contradict one another. They differ in inclusion criteria, comparators, and outcome definitions, and much of the apparent disagreement is a difference in the question being asked.
Fracture healing makes the second point sharper. Peng et al. (2020) pooled 22 randomized trials (1,468 participants) and found a higher healing rate with PEMF (79.7% versus 64.3%; RR = 1.22, 95% CI [1.10, 1.35]), rated as moderate-quality evidence, with very low-quality evidence for faster healing time. Picelli et al. (2024), restricting their update to randomized trials on acute fractures published between 2014 and 2022, found three trials (197 patients), none showing a significant effect on bone healing. Both findings can be true at once. Scope, not only quality, determines what a review can say.
Microcurrent shows the same pattern. In a meta-analysis of randomized trials in wound care, Avendaño-Coy et al. (2022) found that microcurrent added to standard care reduced wound surface area (MD = −8.3 cm², 95% CI [−10.5, −6.0], moderate certainty) and healing time (MD = −7.0 days, 95% CI [−11.9, −2.1], low certainty), but found no difference in the number of wounds fully healed (very low certainty). Iijima and Takahashi (2021) found benefit over sham for knee pain in a single trial of 52 patients and, in the same review, a clinically significant placebo response. A 2026 systematic review of 11 studies found results inconsistent because of variability in protocols, populations, and assessment methods, and concluded that microcurrent should be considered a complementary therapy (Wadhwa et al., 2026).
Every trial cited in this section studied a clinical population. Applied to a person seeking rest rather than treatment, the evidence becomes indirect in GRADE’s sense. For wellness use, then, tier II is a ceiling rather than a floor. Mixed is not a polite word for weak. It is a precise word for heterogeneous.
Tier III: Theoretical biophysics
Entry. A mechanistically coherent model, grounded in credible basic science, that has not been tested in the setting or for the purpose where it is being applied.
Permitted language. Claims explicitly labeled as models or open questions: “researchers are investigating whether…”; “one model proposes…”
Boundary. A tier III claim is never presented as a finding and never priced as a treatment.
Bioelectric signaling is a useful example because the underlying science is strong. Levin (2014) reviews evidence that resting membrane potentials among non-neural cells regulate proliferation, differentiation, and large-scale pattern formation, including regeneration in animal models. Within developmental biology this is serious, well-published work. The inference from it to “a microcurrent session restores the body’s bioelectric blueprint” is tier III at best: the basic science is real, and the bridge from it to a wellness session is a model that has not been tested.
A second example comes from our own practice. In photobiomodulation, the biphasic dose response means that low doses of light can stimulate more effectively than higher ones, so the direction of an effect depends on dose (Huang et al., 2009). Whether low-power visible lasers moving across crystal in an art installation deliver a dose that falls inside any therapeutic window is an open question, because it has not been measured. Until it is, any photobiomodulation claim for the installation sits in tier III, and the installation’s honest claims sit in tier IV.
Below the floor
The gradient has a floor. Claims that contradict well-established physics or biology, or that have failed direct tests, do not enter at tier III as “theories under investigation.” They sit below the gradient and are not made at all. The ALSUntangled Group’s (2014) review of the Rife machine is a clear case: it concluded that Rife’s identification of pathogen-specific frequencies and the ability of Rife machines to kill pathogens “are all unproven and highly implausible” (Conclusion section), and it found no verified case of ALS improving on any objective outcome measure.
This is where rating claims rather than devices does its most important work. A resonant light device offered as a quiet, restful sensory experience can carry an honest tier IV claim. The same device described as destroying pathogens falls below the floor. The device does not change. The claim does.
Tier IV: Experiential wellness art
Entry. The claim concerns subjective, aesthetic, or somatic experience in an immersive setting: calm, awe, presence, rest.
Permitted language. “Many people describe feeling calmer”; “the space is designed for rest.” Physiological language only where physiology has been measured in that setting.
Boundary. No disease claims, and no physiological claims without measurement.
The evidence relevant to this tier is broader than many clinicians expect. Beyond the arts-and-health literature summarized by Fancourt and Finn (2019), a recent trial isolates immersion itself. Nowakowska et al. (2026) delivered identical relaxation content to university students for 10 minutes a day over five consecutive days, either in immersive virtual reality or on a screen. The immersive group showed more favorable changes in heart rate variability and a significantly greater reduction in perceived stress (p = .009, Cohen’s d = 0.78). The authors themselves caution that the intervention period was short, that there was no follow-up, and that adequately powered trials with longer observation are needed.
The counterweight matters as much as the finding. In a within-subjects study of 67 adults on a five-day nature trip, Scott et al. (2021) predicted signs of physiological recovery and found the opposite: respiratory sinus arrhythmia decreased and heart rate increased during the trip, suggesting parasympathetic withdrawal. Self-reported mood improved all the same. Physiological and experiential measures can move in different directions in the same people, in the same setting. That is why tier IV claims are phrased in terms of experience, and why a physiological claim about an immersive space requires measurement in that space.
Tier IV also gives contextual effects an honest place to stand. Iijima and Takahashi (2021) argue that because treatment in real care is always delivered together with a placebo response, and because that response is often clinically meaningful, the overall effect people experience deserves attention alongside the specific effect. Tier IV names that experience openly instead of disguising it as mechanism. It is not a demotion. It is the tier where art is allowed to be art.
Rules of use
- Rate the claim, not the modality. One device can carry a tier I statement, a tier II statement, and a tier IV statement at once. Each is labeled separately.
- Mechanism does not license outcome. A tier I fact supports only tier I language (Howick et al., 2010).
- Discount for indirectness. Clinical evidence applied to wellness use is indirect evidence, and the claim is capped at tier II and stated as such (Guyatt et al., 2008).
- Date the placement. Evidence moves. A placement is a statement about the literature as of a date, and it is revisited when new reviews appear.
- Stay on the wellness side of the line. Wellness claims describe intended use, not demonstrated safety or efficacy (U.S. Food and Drug Administration, 2026).
- Show the work. Every placement carries its citations, so that readers can disagree with a specific placement rather than with the practice as a whole.
Applying the Gradient
The table below shows how claims connected to our own practice are placed, and the wording we use for each.
| Claim | Placement | Basis | Wording we use |
|---|---|---|---|
| A mitochondrial enzyme absorbs red and near-infrared light | Tier I | de Freitas & Hamblin (2016) | “Mitochondria contain an enzyme that absorbs these wavelengths.” |
| PEMF improves function in knee osteoarthritis | Tier II | Chen et al. (2019); Yang et al. (2020); Viganò et al. (2021) | “Some trials in osteoarthritis report benefit; results vary. Our sessions are not a treatment.” |
| PEMF supports fracture healing | Tier II | Peng et al. (2020); Picelli et al. (2024) | Not used in our materials; outside our scope as a wellness practice. |
| Microcurrent speeds wound healing | Tier II | Avendaño-Coy et al. (2022); Wadhwa et al. (2026) | “Clinical wound-care trials show mixed results. We offer microcurrent as a wellness session.” |
| Microcurrent restores the body’s bioelectric blueprint | Tier III | Levin (2014), basic science only | “Bioelectric signaling is an active research field. We do not claim our sessions act on it.” |
| The installation delivers photobiomodulation | Tier III | Huang et al. (2009); dose not measured | “We do not claim a therapeutic light dose.” |
| Frequency devices destroy pathogens | Below the floor | ALSUntangled Group (2014) | Never claimed. |
| Immersive light spaces help people slow down and feel calm | Tier IV | Fancourt & Finn (2019); Nowakowska et al. (2026); Scott et al. (2021) | “Many people describe feeling calmer. The space is designed for rest.” |
Where the Gradient falls short
The framework has real limitations, and a reader trained in evidence appraisal will see most of them quickly.
- It has not been validated. No study has tested whether different raters place the same claims in the same tiers. The placements above are the author’s judgments, supported by citations, not the output of a tested instrument.
- It is applied by an interested party. A practitioner rating claims about her own offerings has an obvious conflict of interest. Publishing the reasoning and the sources reduces that problem without removing it.
- The tiers mix kind and strength. Tier I concerns mechanisms, tier II clinical outcomes, and tier IV experience. A well-replicated tier IV finding may deserve more confidence than a weak tier II one, so the ordering should be read as a sequence of claim types, not as a strict ranking of certainty. A two-axis version, with claim type on one axis and certainty on the other, may describe the evidence more faithfully.
- It imposes categories on a continuum. This is the arbitrariness Guyatt et al. (2008) acknowledge for GRADE, and it applies here with at least equal force.
- It can be misused. Any labeling scheme can become a veneer, lending an air of rigor to claims that have not earned it. The safeguard is the rule to show the work.
A research agenda
Each weakness above suggests a study.
- Inter-rater reliability. Give a set of real wellness claims to practitioners and to health researchers, ask each to place them on the gradient, and measure agreement.
- Audience comprehension. Test whether tier labels change how members of the public understand and trust wellness claims, compared with the same claims unlabeled.
- Session-level measurement. Measure heart rate variability alongside validated self-report within immersive installations, so that tier IV physiological claims can be tested in the setting where they are made rather than borrowed from virtual reality or nature studies.
- Dose measurement. Measure irradiance at the participant’s position in a light installation and compare it with the dose ranges reported in photobiomodulation research, which would move the installation’s photobiomodulation question out of tier III one way or the other.
The first three involve human participants and would require institutional ethical review before any data collection. The fourth does not, and it is the natural place to begin.
Why label at all
Wellness will keep producing confident claims, and much of what it offers people is real, even when it is not what the marketing says. The Evidence Gradient does not ask anyone to stop exploring. It asks that exploration be labeled, so that a reader can tell a mechanism from a trial, a trial from a model, and a model from an experience. Readers who work with evidence frameworks professionally are invited to challenge any placement here. As the literature moves, the placements will move with it.
The device does not change. The claim does.
Where would you place the claims made in your own field?
References
ALSUntangled Group. (2014). ALSUntangled No. 23: The Rife machine and retroviruses. Amyotrophic Lateral Sclerosis and Frontotemporal Degeneration, 15(1–2), 157–159. https://doi.org/10.3109/21678421.2013.850802
Avendaño-Coy, J., López-Muñoz, P., Serrano-Muñoz, D., Comino-Suárez, N., Avendaño-López, C., & Martin-Espinosa, N. (2022). Electrical microcurrent stimulation therapy for wound healing: A meta-analysis of randomized clinical trials. Journal of Tissue Viability, 31(2), 268–277. https://doi.org/10.1016/j.jtv.2021.12.002
Chen, L., Duan, X., Xing, F., Liu, G., Gong, M., Li, L., Chen, R., & Xiang, Z. (2019). Effects of pulsed electromagnetic field therapy on pain, stiffness and physical function in patients with knee osteoarthritis: A systematic review and meta-analysis of randomized controlled trials. Journal of Rehabilitation Medicine, 51(11), 821–827. https://doi.org/10.2340/16501977-2613
de Freitas, L. F., & Hamblin, M. R. (2016). Proposed mechanisms of photobiomodulation or low-level light therapy. IEEE Journal of Selected Topics in Quantum Electronics, 22(3), 348–364. https://doi.org/10.1109/JSTQE.2016.2561201
Fancourt, D., & Finn, S. (2019). What is the evidence on the role of the arts in improving health and well-being? A scoping review (Health Evidence Network Synthesis Report No. 67). WHO Regional Office for Europe. https://www.ncbi.nlm.nih.gov/books/NBK553773/
Guyatt, G. H., Oxman, A. D., Vist, G. E., Kunz, R., Falck-Ytter, Y., Alonso-Coello, P., & Schünemann, H. J. (2008). GRADE: An emerging consensus on rating quality of evidence and strength of recommendations. BMJ, 336(7650), 924–926. https://doi.org/10.1136/bmj.39489.470347.AD
Howick, J., Glasziou, P., & Aronson, J. K. (2010). Evidence-based mechanistic reasoning. Journal of the Royal Society of Medicine, 103(11), 433–441. https://doi.org/10.1258/jrsm.2010.100146
Huang, Y.-Y., Chen, A. C.-H., Carroll, J. D., & Hamblin, M. R. (2009). Biphasic dose response in low level light therapy. Dose-Response, 7(4), 358–383. https://doi.org/10.2203/dose-response.09-027.Hamblin
Iijima, H., & Takahashi, M. (2021). Microcurrent therapy as a therapeutic modality for musculoskeletal pain: A systematic review accelerating the translation from clinical trials to patient care. Archives of Rehabilitation Research and Clinical Translation, 3(3), Article 100145. https://doi.org/10.1016/j.arrct.2021.100145
Levin, M. (2014). Molecular bioelectricity: How endogenous voltage potentials control cell behavior and instruct pattern regulation in vivo. Molecular Biology of the Cell, 25(24), 3835–3850. https://doi.org/10.1091/mbc.E13-12-0708
Murad, M. H., Asi, N., Alsawas, M., & Alahdab, F. (2016). New evidence pyramid. Evidence-Based Medicine, 21(4), 125–127. https://doi.org/10.1136/ebmed-2016-110401
Nowakowska, A., Nowak, M., Bigas, J., Gaweł, Z., Nęcki, M., & Rutkowski, S. (2026). Short-term psychophysiological effects of immersive versus non-immersive relaxation in university students: A randomized controlled trial. Healthcare, 14(15), Article 2322. https://doi.org/10.3390/healthcare14152322
Peng, L., Fu, C., Xiong, F., Zhang, Q., Liang, Z., Chen, L., He, C., & Wei, Q. (2020). Effectiveness of pulsed electromagnetic fields on bone healing: A systematic review and meta-analysis of randomized controlled trials. Bioelectromagnetics, 41(5), 323–337. https://doi.org/10.1002/bem.22271
Picelli, A., Di Censo, R., Tomasello, S., Scaturro, D., Letizia Mauro, G., Smania, N., & Filippetti, M. (2024). Effects of pulsed electromagnetic fields on bone fractures: A systematic review update. European Journal of Physical and Rehabilitation Medicine, 60(6), 989–994. https://doi.org/10.23736/S1973-9087.24.08226-1
Russo, F., & Williamson, J. (2007). Interpreting causality in the health sciences. International Studies in the Philosophy of Science, 21(2), 157–170. https://doi.org/10.1080/02698590701498084
Scott, E. E., LoTemplio, S. B., McDonnell, A. S., McNay, G. D., Greenberg, K., McKinney, T., Uchino, B. N., & Strayer, D. L. (2021). The autonomic nervous system in its natural environment: Immersion in nature is associated with changes in heart rate and heart rate variability. Psychophysiology, 58(4), Article e13698. https://doi.org/10.1111/psyp.13698
U.S. Food and Drug Administration. (2026). General wellness: Policy for low risk devices: Guidance for industry and Food and Drug Administration staff. https://www.fda.gov/media/90652/download
Viganò, M., Perucca Orfei, C., Ragni, E., Colombini, A., & de Girolamo, L. (2021). Pain and functional scores in patients affected by knee OA after treatment with pulsed electromagnetic and magnetic fields: A meta-analysis. Cartilage, 13(1 Suppl.), 1749S–1760S. https://doi.org/10.1177/1947603520931168
Wadhwa, K., Singh, A. K., Bhatnagar, R. B., & Kapoor, K. (2026). Role of microcurrent electrical stimulation in tissue healing and scar modulation: A systematic review. The International Journal of Lower Extremity Wounds. Advance online publication. https://doi.org/10.1177/15347346261469606
Yang, X., He, H., Ye, W., Perry, T. A., & He, C. (2020). Effects of pulsed electromagnetic field therapy on pain, stiffness, physical function, and quality of life in patients with osteoarthritis: A systematic review and meta-analysis of randomized placebo-controlled trials. Physical Therapy, 100(7), 1118–1131. https://doi.org/10.1093/ptj/pzaa054