References

Exercise as a mental health intervention

Noetel et al., The BMJ, 2024. Network meta-analysis, 218 trials, 495 arms, 14,170 people: strength training, walking or jogging, and yoga reduced depression symptoms as effectively as cognitive behavioural therapy. doi:10.1136/bmj-2023-075847
Munro et al., British Journal of Sports Medicine, 2026. Umbrella review with meta-meta-analysis, 63 reviews covering 1,079 trials and 79,551 participants: depression SMD −0.61, anxiety −0.47, with aerobic exercise showing the largest effect. doi:10.1136/bjsports-2025-110301
Díaz-Goñi et al., Nature Mental Health, 2026. Meta-analysis, 27 studies, 4,007,638 people across nine countries: high cardiorespiratory fitness went with a 36% lower risk of depression, 39% lower dementia and 29% lower psychotic disorders. One extra MET went with about 5% lower depression risk and 19% lower dementia risk. doi:10.1038/s44220-026-00599-4
Lang et al., British Journal of Sports Medicine, 2024. Overview of 26 systematic reviews, 20.9 million observations from 199 cohort studies: cardiorespiratory fitness is a strong and consistent predictor of illness and early death. doi:10.1136/bjsports-2023-107849
Toval et al., European Heart Journal, 2025. Network meta-analysis in coronary artery disease: exercise improved quality of life, depression and anxiety, and in-person programmes did better than home-based ones. doi:10.1093/eurheartj/ehae870

Addiction and recovery

Liu, Chen & Gao, Frontiers in Psychiatry, 2026. Network meta-analysis, 57 randomised trials, 2,922 participants: exercise improved substance-use outcomes, mental health, cognition, physiological health and sleep. Aerobic exercise was best for physiological health, mind-body exercise for sleep. doi:10.3389/fpsyt.2025.1732663
Wang, Wang, Wang, Li & Zhou, PLOS ONE, 2014. Meta-analysis of 22 studies: exercise raised abstinence rates (OR 1.69) and eased withdrawal symptoms. The abstinence result is the solid one — the withdrawal effect only just cleared significance, so treat it as suggestive. doi:10.1371/journal.pone.0110728
Robertson et al., Neuropsychopharmacology, 2016. UCLA trial: eight weeks of treadmill walking or jogging three times a week plus resistance training raised striatal D2/D3 receptor availability by about 15% in methamphetamine users. A small study — 10 exercisers and 9 controls. doi:10.1038/npp.2015.331

Heart rate variability and the autonomic nervous system

Wang et al., Translational Psychiatry, 2025. Umbrella review of meta-analyses: heart rate variability in mental disorders. Lower HRV tracks with anxiety and depression. doi:10.1038/s41398-025-03339-x
Wendt & Thayer, ‘Heart Rate Variability in Mental Health and Psychotherapy’, chapter 4 in Steffen & Moss (eds.), Integrating Psychotherapy and Psychophysiology, Oxford University Press, 2024. A review of the HRV–mental health association and the neurovisceral integration model behind it. A book chapter, not a peer-reviewed paper — the Wang umbrella review above is the stronger citation for the same point. doi:10.1093/oso/9780198888727.003.0004
Presby et al., Journal of Medical Internet Research, 2025. Thirteen months of wearable data from 181,574 people: lower resting heart rate, higher HRV and better sleep tracked with lower depression, anxiety and stress. Note one wrinkle — longer sleep went with better mental health between people, but within a single person an increase in sleep duration went with higher depression. doi:10.2196/64955
Satapathy et al., Brain and Behavior, 2025. Meta-analysis of 36 studies: depression linked to roughly twenty percent higher odds of hypertension (OR 1.20), with a pooled hazard ratio of 1.28 in the longitudinal studies. doi:10.1002/brb3.70931
Kral et al., Scientific Reports, 2023. In 245 people after wellness training, a slower respiration rate went with higher self-reported well-being. doi:10.1038/s41598-023-43176-w
Manresa-Rocamora et al., Clinical Autonomic Research, 2021 (online 2020). Meta-analysis: exercise-based cardiac rehabilitation improved parasympathetic measures in coronary artery disease — RMSSD and heart rate recovery both improved, though high-frequency power did not. doi:10.1007/s10286-020-00687-0
Manresa-Rocamora et al., International Journal of Environmental Research and Public Health, 2021. Meta-analysis of HRV-guided training: it improved vagal-related HRV more than standard training. It did not significantly improve aerobic capacity or endurance performance — the gains there were small and did not reach significance, so this supports guiding training by HRV for autonomic reasons, not for performance. doi:10.3390/ijerph181910299
Grossman et al., Clinical Neuropsychiatry, 2026. Thirty-nine experts in vagal physiology and vertebrate social behaviour conclude that the major tenets of polyvagal theory are not supported by current knowledge. doi:10.36131/cnfioritieditore20260110 — with Porges’s reply in the same issue, doi:10.36131/cnfioritieditore20260111
Grossman, Biological Psychology, 2023. The earlier, single-author version of the same critique: challenges to the five basic premises of polyvagal theory. doi:10.1016/j.biopsycho.2023.108589

Nature-based interventions

Saeedy Robat et al., Nature Human Behaviour, 2026. Overview of 116 systematic reviews, with a second-order meta-analysis of 30 of them covering 3,870 primary studies: nature-based interventions reduced stress, anxiety and depression. The often-quoted “ten million participants” is an estimate pooled across those reviews, not ten million people enrolled in one study. The same paper found the largest benefits from multi-day immersive programmes and those led by people with clinical or outdoor expertise. doi:10.1038/s41562-026-02433-4
Liang, Wei & Min, Frontiers in Public Health, 2026. Network meta-analysis ranking green exercise above indoor exercise for depression and anxiety. A small evidence base — 15 randomised trials and 809 participants — so treat it as promising rather than settled. doi:10.3389/fpubh.2026.1831073
Bettmann et al., International Journal of Mental Health Nursing, 2024. Meta-analysis: nature exposure produced moderate improvements in adults who already had symptoms of mental illness, with smaller but still significant effects in the controlled studies alone. doi:10.1111/inm.13400
Siah et al., International Journal of Mental Health Nursing, 2023. Meta-analysis of forest bathing: significant reductions in depression and anxiety. The physiological effects were much weaker than the psychological ones — the benefit here is to how people feel, not to their biology. doi:10.1111/inm.13131
Wang, Feng & Wang, Frontiers in Public Health, 2025. Umbrella review of 47 meta-analyses on green and blue space: lower all-cause and cardiovascular mortality, less diabetes and metabolic syndrome, better mental health. The authors rate most of the underlying evidence as observational and low-quality, and the cancer and asthma findings as inconclusive. Association, not proof of cause. doi:10.3389/fpubh.2025.1505292

Psychological screening

Kroenke, Spitzer & Williams, Journal of General Internal Medicine, 2001. The PHQ-9: nine items, depression severity over two weeks. A score of 10 or more indicates at least moderate depression, with 88% sensitivity and 88% specificity for major depression. doi:10.1046/j.1525-1497.2001.016009606.x
Kroenke, Spitzer, Williams & Löwe, General Hospital Psychiatry, 2010. Systematic review of the whole PHQ family — depression, anxiety and somatic scales — confirming the cutoff of 10 across them. doi:10.1016/j.genhosppsych.2010.03.006
Spitzer, Kroenke, Williams & Löwe, Archives of Internal Medicine, 2006. The GAD-7: seven items, generalised anxiety. At a cutoff of 10, sensitivity and specificity both exceed 0.80. doi:10.1001/archinte.166.10.1092
Lovibond & Lovibond, Behaviour Research and Therapy, 1995. The original Depression Anxiety Stress Scales. This is the 42-item DASS, not the DASS-21 — the short form comes from Henry & Crawford below, and the two are often confused. doi:10.1016/0005-7967(94)00075-U
Henry & Crawford, British Journal of Clinical Psychology, 2005. The DASS-21: the twenty-one item short form, validated in a large non-clinical sample. The three subscales are valid, but each also picks up a general dimension of distress. doi:10.1348/014466505X29657
Graham et al., Annals of Family Medicine, 2019. Validation of the CAT-MH computerised adaptive test in 271 primary care patients, against the PHQ-9, PHQ-2 and GAD-7. doi:10.1370/afm.2316
Kraska, Bell & Costello, Journal of Medical Internet Research, 2023. Graded response model analysis and computer adaptive test simulation of the DASS-21. Its items work best in the mild-to-moderate range. doi:10.2196/45334

Sleep

Scott et al., Sleep Medicine Reviews, 2021. Meta-analysis of 65 randomised trials, 8,608 participants: improving sleep quality improved mental health overall (g = −0.53), depression (−0.63), anxiety (−0.51) and stress (−0.42). doi:10.1016/j.smrv.2021.101556

Nutrition, the gut, and the vagus nerve

Aburto & Cryan, Nature Reviews Gastroenterology & Hepatology, 2024. Review of how signals from gut microbes cross the gut lining, the blood-brain barrier and the blood-CSF barrier. doi:10.1038/s41575-023-00890-0
Schneider, O’Riordan, Clarke & Cryan, Nature Metabolism, 2024. Review of diet as the lever on the gut-brain axis: what you eat reaching the brain by way of the gut. doi:10.1038/s42255-024-01108-6
Bonaz, Bazin & Pellissier, Frontiers in Neuroscience, 2018. The standard reference for the vagus nerve’s place in the microbiota-gut-brain axis. doi:10.3389/fnins.2018.00049
Hall et al., Cell Metabolism, 2019. Inpatient randomised trial: on an ultra-processed diet people ate about 500 calories a day more, and gained weight, compared with a matched unprocessed diet. The foundational experiment here. doi:10.1016/j.cmet.2019.05.008
Dicken et al., Nature Medicine, 2025. Randomised crossover trial, 55 adults, two eight-week diets both built to meet national healthy-eating guidance — one minimally processed, one ultra-processed. The minimally processed diet produced more weight loss even though both met the guidelines. doi:10.1038/s41591-025-03842-0
Brunstrom et al., American Journal of Clinical Nutrition, 2026. Re-analysis of the Hall trial: on the unprocessed diet people built meals with a less balanced carbohydrate-to-fat mix, which predicted smaller meals and lower intake. A post-hoc analysis of 20 people — a mechanism worth knowing about, not a settled one. doi:10.1016/j.ajcnut.2025.101183

On the vagus nerve and inflammation, the honest position. The idea that stimulating the vagus nerve calms inflammation is well established in rodents and is not consistently demonstrated in people. The best available evidence is a null result, and it is listed here rather than left out:

Schiweck et al., Brain, Behavior, and Immunity, 2024. Pre-registered systematic review and meta-analysis: no consistent evidence for an anti-inflammatory effect of vagus nerve stimulation in humans. doi:10.1016/j.bbi.2023.12.008
de Melo et al., Neuromodulation, 2025. Meta-analysis, 15 studies, 597 patients: no significant pooled effect on TNF-α, IL-6 or IL-1β; some effect on CRP, IL-10 and IFN-γ. A mixed, partial result. doi:10.1016/j.neurom.2024.03.002

Performance-enhancing drugs, peptides and HRV

Maior et al., Scandinavian Journal of Medicine & Science in Sports, 2013. Twenty-two recreationally trained men, eleven using anabolic steroids: lower resting HRV across SDNN, RMSSD, pNN50 and spectral power, and lower parasympathetic measures after exercise. doi:10.1111/j.1600-0838.2011.01436.x
Kouidi et al., International Journal of Environmental Research and Public Health, 2021. Forty steroid-using strength athletes against forty non-users: baroreflex sensitivity lower by 13.8%, sympathetic activity higher, and left-ventricular diastolic changes that correlated with years of use. doi:10.3390/ijerph18136974
Kumarathurai et al., Diabetes Care, 2017. Randomised, double-blind, placebo-controlled crossover trial: liraglutide raised mean heart rate by 8.1 beats a minute and lowered HRV, with SDNN down 33.9 ms. doi:10.2337/dc16-1580
Grosicki et al., American Journal of Physiology – Heart and Circulatory Physiology, 2025. Sixty-six people on a wearable: over twelve weeks on a GLP-1 drug, resting heart rate rose 3.2 beats a minute, mediated by a 6.2 ms drop in HRV. doi:10.1152/ajpheart.00809.2024
Simon et al., Obstetrics & Gynecology, 2019. Fifty-two week open-label extension of the bremelanotide phase 3 trials: nausea 40.4%, flushing 20.6%, headache 12.0%. doi:10.1097/AOG.0000000000003514
U.S. Food and Drug Administration, Certain Bulk Drug Substances for Use in Compounding that May Present Significant Safety Risks. The FDA placed BPC-157 in Category 2 of this list in 2023, citing possible immunogenicity, peptide-related impurities and limited human safety data. It removed BPC-157 from Category 2 in April 2026, after the nomination was withdrawn. That removal is not a safety clearance: the FDA says it does not make a substance eligible for compounding, and the substance remains under evaluation.
Health Canada, “Think twice before injecting peptides bought online: unauthorized products can seriously harm you”, 9 April 2026. Peptides are generally regulated as prescription drugs in Canada. Unauthorised products have not been assessed for safety, efficacy or quality, and the advisory lists the harms.
Pietilä et al., JMIR Mental Health, 2018. 4,098 Finnish employees measured on wearables: alcohol lowered HRV-derived recovery in the first hours of sleep by 9.3, 24.0 and 39.2 percentage units at low, moderate and high intake. doi:10.2196/mental.9519

How the sensors work, and how good they are

The lights on the back of a watch, what they measure and how closely it agrees with a medical ECG. Written up on the green lights on the back of your watch.

Xu, Liu, Liu & Su, Sensors, 2026;26(16):5192. Systematic review and meta-analysis of 43 studies: pooled absolute standardised error against ECG of 0.188 for RMSSD and 0.134 for SDNN in healthy adults. doi:10.3390/s26165192
Coste, Millour & Hausswirth, Sensors, 2025;25(18):5745. Posture changes the answer. Agreement with ECG over five minutes: lying down ICC 0.955 for RMSSD and 0.980 for SDNN; seated 0.834 and 0.921, with wider limits of agreement. This is why the morning reading is taken lying down. doi:10.3390/s25185745
Bellenger, Miller, Halson, Roach & Sargent, Sensors, 2021;21(10):3571. Wrist PPG against ECG during slow-wave sleep: correlations of 0.98–0.99 for RMSSD. The authors also note the bias and limits of agreement in ln RMSSD approached or exceeded the smallest worthwhile change, which is a real caution for reading small day-to-day differences. doi:10.3390/s21103571
van Dijk, Huizink, Oosterman & Lemmers-Jansen, Psychosomatic Medicine, 2023;85(7):568–576. Validation of phone-camera photoplethysmography for HRV, the method the morning-reading apps use. doi:10.1097/PSY.0000000000001236
Bent, Goldstein, Kibbe & Dunn, npj Digital Medicine, 2020. Where wearable optical heart rate sensors go wrong, and why. doi:10.1038/s41746-020-0226-6
Zhang et al., Sensors, 2019;19(3):673. Green versus infrared, and using the deeper infrared channel as a movement reference — heart rate error in motion cut from about 4 bpm to about 1. doi:10.3390/s19030673
Koerber, Khan, Shamsheri, Kirubarajan & Mehta, Journal of Racial and Ethnic Health Disparities, 2022. Ten studies on accuracy across skin tones: four found reduced accuracy in darker skin tones, four found no difference, two were mixed. Unsettled, and not nothing. doi:10.1007/s40615-022-01446-9
Muggeridge et al., JMIR mHealth and uHealth, 2021. Heart rate measurement with the Polar OH1 and Fitbit Charge 3 across activities. doi:10.2196/25313

Where you sit — Canadian fitness norms

The population comparison in an assessment report uses Canadian data, not American. These are the reference sets.

Hoffmann, Colley, Doyon, Wong, Tomkinson & Lang, Health Reports (Statistics Canada), 2019;30(10):14–22. Nationally representative percentile values for five fitness tests, by age group and sex, from over 5,000 Canadians aged 3 to 79 in the Canadian Health Measures Survey. This is the table a Canadian percentile should come from. doi:10.25318/82-003-x201901000002-eng
Wong, Health Reports (Statistics Canada), 2016;27(10):3–10. Grip strength reference values for Canadians aged 6 to 79, from the Canadian Health Measures Survey 2007–2013. statcan.gc.ca
Trends in physical fitness among Canadian adults, 2007 to 2017, Health Reports (Statistics Canada), 2021. What has happened to Canadian fitness over a decade, which is the context a single percentile lacks. statcan.gc.ca
CSEP-PATH Resource Manual, 3rd edition, Canadian Society for Exercise Physiology, September 2021. The Canadian professional standard for health-related fitness assessment protocols and their interpretation. store.csep.ca

How current this actually is, stated plainly. The newest published Canadian fitness percentiles are the 2019 paper above, built on Canadian Health Measures Survey cycle 5. The survey’s latest cycle ran to December 2024, but its first release, on 17 October 2025, covered physical activity and sedentary time only — not VO2 max, grip strength or the other fitness measures. So there is no 2025 Canadian fitness norm to use yet. When one is published, this section and the assessment reports should be updated together.

Grip strength

Handgrip dynamometry is used in the assessment and in the mobile metabolic laboratory. It is the cheapest measure taken and the one with the largest evidence base behind it.

Leong, Teo, Rangarajan et al., The Lancet, 2015;386(9990):266–273. The Prospective Urban Rural Epidemiology (PURE) study: 139,691 participants across 17 countries. Each 5 kg lower grip strength was associated with a 16% higher risk of all-cause mortality (HR 1.16, 95% CI 1.13–1.20) and a 17% higher risk of cardiovascular mortality (HR 1.17, 1.11–1.24). Grip strength predicted all-cause death better than systolic blood pressure did. doi:10.1016/S0140-6736(14)62000-6
Cabanas-Sánchez, Esteban-Cornejo, Parra-Soto et al., Journal of Cachexia, Sarcopenia and Muscle, 2022;13(4):1983–1994. UK Biobank, 162,167 participants aged 38 to 70. Each 5 kg lower grip strength was associated with a 7% higher risk of depression and an 8% higher risk of anxiety. This is the closest thing to direct population evidence for the link Resilience Therapy is built on. doi:10.1002/jcsm.12963
Wong, Health Reports (Statistics Canada), 2016 — the Canadian percentiles, listed above under Canadian fitness norms. Measured standing, twice on each hand alternating, dynamometer in line with the forearm at thigh level. A reading taken seated with the elbow at ninety degrees is a different number and cannot be read off the same table.

What is not settled. Whether strength protects mood, low mood costs strength, or both follow something else again, is unresolved. The association is large and repeatedly replicated; the direction of cause is not established, and nothing on this site should say it is.

Strength, ageing and injury

Lopez et al., Obesity Reviews, 2022. Systematic review and meta-analysis of resistance training across the lifespan in people with overweight and obesity: effects on body composition and body weight. doi:10.1111/obr.13428
Cunha et al., Psychiatry Research, 2024. Systematic review and meta-analysis of randomised trials: can resistance training improve mental health outcomes in older adults. doi:10.1016/j.psychres.2024.115746
Radaelli et al., Sports Medicine, 2025. Network meta-analysis of 151 randomised trials on resistance training volume, physical function, lean mass and strength in older adults. doi:10.1007/s40279-024-02123-z

Boxing and combat sports

Bozdarov, Jones, Daskalakis & Husain, American Journal of Lifestyle Medicine, 2022. Scoping review of sixteen documents using non-contact boxing as a mental health intervention: most reported reduced stress and improved mood, self-esteem and quality of life. A scoping review maps what has been studied; it does not establish effect sizes. doi:10.1177/15598276221124095
Ciaccioni et al., Psychology of Sport and Exercise, 2024. Systematic review of martial arts, combat sports and mental health in adults. doi:10.1016/j.psychsport.2023.102556
Bozdarov et al., PLOS ONE, 2025. Feasibility study of mindfulness-based non-contact boxing therapy for depression and anxiety. A feasibility study — designed to show the thing can be run, not that it works. doi:10.1371/journal.pone.0318364
Albaladejo-García et al., Neuroscience & Biobehavioral Reviews, 2023. Systematic review and meta-analysis of the stop-signal paradigm: athletes stopped an action already under way faster than non-athletes (SMD 0.44). Inhibition is the executive function closest to what combat sports train. doi:10.1016/j.neubiorev.2023.105108
Contreras-Osorio et al., International Journal of Environmental Research and Public Health, 2022. Meta-analysis of nine sport-based training trials in older adults, 398 participants: working memory improved (ES 0.35). Inhibition and cognitive flexibility moved the same way but did not reach significance. doi:10.3390/ijerph191912573