This article is part of the HBOT Radar series, where we summarize the latest published hyperbaric oxygen therapy research.
Fresh Science, Clearly Explained — Brought to you by Brain Spa Hyperbaric
Disclaimer: This article is intended for educational and informational purposes only. It summarizes published medical research conducted in clinical settings and does not evaluate Brain Spa Hyperbaric products. The hyperbaric chambers offered on this website are non-medical wellness devices and are not intended to diagnose, treat, cure, or prevent any disease. Do not make medical decisions based on this article — consult a qualified healthcare professional.
📌 9 in 10 Parkinson's Patients Improved. Nobody Recorded the Dose.
🔍 Why this topic matters
Parkinson's disease is a slow subtraction. It begins with dopamine-producing neurons dying in a small region of the midbrain called the substantia nigra, and it works outward from there — tremor, rigidity, slowed movement, difficulty walking. In advanced stages, many patients become bedridden.
Levodopa and deep brain stimulation can hold the motor symptoms at bay for a while. Neither slows what's actually happening underneath.
There's a reason researchers keep returning to oxygen. The authors of this paper make the case directly: nearly every known risk factor for Parkinson's — genetics, ageing, environmental exposure, brain injury — shares one feature, in that they all impair the brain's ability to take in and use oxygen. Genetic factors damage mitochondria, the cell's oxygen-processing machinery. Ageing reduces oxygen utilisation across the body. And low oxygen has been shown to accelerate the abnormal clumping of alpha-synuclein, the protein that drives Parkinson's pathology and kills dopamine neurons. It's a coherent argument rather than a proven one, and it's the reasoning behind a growing body of work — including the animal studies on HBOT in Alzheimer's and Parkinson's we covered separately.
If the disease involves a brain that can't get or use enough oxygen, the obvious question is whether delivering more helps. A team at Tongji University School of Medicine in Shanghai pooled every randomized controlled trial they could find to answer it.
🔬 What the researchers reviewed
A systematic review and meta-analysis searching five databases — PubMed, Web of Science, Cochrane Library, and the Chinese databases CNKI and Wanfang — through March 2023.
Thirteen randomized controlled trials with 958 participants met the criteria: 496 receiving hyperbaric oxygen therapy alongside standard Parkinson's medication, 462 receiving medication alone. Trials were published between 1995 and 2022, all conducted in China.
Control groups received levodopa (madopar) or dopamine agonists (pramipexole). In one trial the standard regimen also included neurotrophic drugs, and in two others it included piracetam, nimodipine, vinpocetine and verapamil — drugs aimed at brain metabolism and cerebral blood flow.
The protocols, and the hole in them. Where pressure was reported, it ranged from 2.0 to 2.5 ATA. Two of the thirteen trials didn't report pressure at all. Four didn't report oxygen concentration; the nine that did all used 100%.
Most treatment courses ran 8 weeks, with the full range spanning one week to three months. Session length and weekly frequency were not reported by any trial.
So the dose these 958 patients actually received is unknown. Eight weeks of treatment could mean sixteen sessions or forty. And that gap matters more than it first appears — though not in the direction most people assume.
Why "more pressure" is not the safe assumption
The instinct when a dose goes unrecorded is to worry that patients were undertreated. In neurological conditions, the more interesting possibility runs the other way.
Paul Harch has spent three decades treating brain-injured patients with hyperbaric oxygen, and he argues that the dose-response curve here is not a ramp but a bell. Too little does nothing. Too much also does nothing — and past a point, it does harm.
He has the clinical record to make that concrete. In a 2024 review of his own practice spanning 1989 to 2023, he documented 59 cases of central nervous system oxygen toxicity occurring below 2.0 ATA — pressures the textbooks had long treated as too low for toxicity to be possible at all. The pattern tracked cumulative exposure rather than pressure alone. In the chronic cases — 52 of the 59 — symptoms typically appeared somewhere around 100 to 120 accumulated atmosphere-hours.
That unit is worth a sentence, because it's the whole point. An atmosphere-hour is the pressure multiplied by the time spent at it, added up across every session a person has had. One hour at 1.5 ATA is one and a half atmosphere-hours. Ninety minutes at 2.0 ATA is three. So the same threshold arrives after roughly seventy sessions at the gentler setting — but after only about thirty-seven at the harder one. The dose isn't the pressure, and it isn't the number of sessions. It's both, multiplied, and it accumulates.
None of the trials pooled here reported how long their sessions lasted, so where these 958 patients landed on that scale is unknowable.
Reading across trials in traumatic brain injury, stroke, cerebral palsy and PTSD, Harch's conclusion is that moderate pressures around 1.3 to 1.5 ATA have repeatedly matched or outperformed 2.0 to 2.4 ATA. His framing is his own and it is contested — a single clinician's interpretation, built partly on a dose metric he devised himself. But it points at something the field mostly skips past. When a hyperbaric trial in a neurological condition comes back null, the standard reading is that the therapy doesn't work. An equally available reading is that the dose landed on the far side of the curve.
Every pressure reported in this meta-analysis sat at 2.0 ATA or above. Nobody tested lower. Whether that helped these results or held them back cannot be answered from this evidence — and it is precisely the question the next generation of trials should be built to answer.
Outcomes measured: overall treatment response (Webster scale), motor function (UPDRS Part III), disease progression (Hoehn-Yahr staging), sleep quality and daytime sleepiness, depression, and cognitive function.
These studies were conducted in clinical hospital settings under medical supervision. Results from clinical research cannot be directly applied to other settings or devices — always consult a healthcare professional.
📊 What the evidence shows
The clearest number: 91% improved versus 76%
Four trials assessed overall treatment response using the Webster scale, a standard Parkinson's rating where an 11% or greater improvement in total score counts as a meaningful response.
90.91% of patients receiving HBOT alongside medication improved, compared with 76.19% on medication alone — odds ratio 3.18 (95% CI: 1.60–6.33, p < 0.01).
What makes this the strongest result in the analysis isn't the size of the gap. It's that the four trials agreed completely with each other (I² = 0%). Four separate research teams, working at different times, arrived at the same answer. In pooled evidence, that kind of convergence is about as good as it gets.
Motor symptoms: a difference at the edge of noticeable
The Unified Parkinson's Disease Rating Scale Part III is what a neurologist fills in while watching you move. They score tremor, rigidity, how quickly you can tap your fingers, whether your face still shows expression, how you rise from a chair, how you walk and turn. Around thirty items, each scored 0 to 4, adding up to a scale that runs from 0 to over 100. Higher means worse.
Across nine trials, HBOT groups scored 2.96 points lower than controls (95% CI: −4.31 to −1.61, p < 0.01).
Researchers have spent years working out how much movement on this scale a patient actually feels. The answer lands somewhere between 2.5 and 5 points, depending on the method used. So 2.96 sits at the bottom edge of noticeable: the size of change where someone might say buttoning a shirt has got a little easier, not the size where a family feels they've got someone back.
The other caveat is that the trials disagreed substantially on the size of the effect (I² = 77%). The direction was consistent; the magnitude was not. Something differed between these studies — patient severity, disease duration, protocol, or assessment practice — that the analysis can't identify.
Disease staging shifted — roughly one patient in seven
Hoehn-Yahr staging is the oldest and simplest way of describing how far Parkinson's has advanced. It isn't a scan or a blood test — it's a clinician's judgement based on what they can observe. Stage 1 means symptoms on one side of the body only. Stage 2 means both sides, balance still intact. Stage 3 is the line that changes a life: balance is now affected, though the person remains independent. Stage 4 means still able to stand and walk, but needing substantial help. Stage 5 is wheelchair-bound or bedridden.
Across five trials, HBOT groups scored 0.14 points lower than controls (95% CI: −0.26 to −0.02, p < 0.01), with all five trials in complete agreement (I² = 0%).
An average of 0.14 sounds like nothing, and for any individual it would be — you can't move a seventh of a stage. What it describes is group arithmetic. A shift of that size is roughly what you'd see if about one patient in seven moved down a full stage while everyone else stayed exactly where they were. One person in seven going from needing help with their balance to not needing it.
One caution about what this measures. Because Hoehn-Yahr is scored from observed symptoms rather than from biology, a treatment that improves symptoms can move the staging without touching the disease underneath. It sits closer to progression than a pure symptom score does, but eight weeks is far too short to say anything about disease modification.
Cognition improved consistently across trials
Three trials measured cognitive function using MoCA and MMSE, the two standard tests for tracking memory, attention, and thinking in neurological disease.
The pooled effect was a standardised mean difference of 0.65 favouring HBOT (95% CI: 0.45–0.85, p < 0.01), with complete agreement across the studies (I² = 0%). An effect size of 0.65 is moderate-to-large — comparable to what approved cognitive drugs achieve in their own trials.
It's worth reading alongside the separate meta-analysis of 847 Alzheimer's patients, which found a similar direction of effect on cognition in a different neurodegenerative population — and which shares many of the same methodological weaknesses.
Daytime sleepiness: back inside the normal range
Excessive daytime sleepiness is one of the most disabling features of Parkinson's. Patients fall asleep mid-conversation, mid-meal, sometimes at the wheel.
It's measured on the Epworth Sleepiness Scale, which asks how likely you are to doze off in eight ordinary situations — reading, watching television, sitting in traffic. The scale runs 0 to 24. Under 10 is normal. Above 10 means excessive daytime sleepiness. The high teens describe someone genuinely struggling to stay awake through a normal day.
Across five trials, HBOT groups scored 3.30 points lower than controls (95% CI: −3.77 to −2.83, p < 0.01), again with zero disagreement between studies (I² = 0%).
On a 24-point scale that is a real move. Someone arriving at 13 — nodding off during conversations, unable to sit quietly after lunch without drifting — leaves at around 9.7, back inside the normal range. That's the kind of change a person and the people around them notice without being told to look for it.
Sleep quality improved — with one thing to hold loosely
Five trials measured sleep quality using the Pittsburgh Sleep Quality Index, a questionnaire covering how long it takes to fall asleep, how often you wake in the night, how rested you feel, and whether you need medication to sleep at all. It runs 0 to 21, and anything above 5 marks a poor sleeper. Parkinson's patients with sleep complaints typically score in the teens.
HBOT groups scored 2.52 points better than controls (95% CI: −2.85 to −2.18, p < 0.01) — enough to move someone from around 13 to 10.5. Real improvement, though not enough on its own to turn a poor sleeper into a good one.
This is the one result to hold more loosely than the others. The paper describes the five trials as showing no meaningful disagreement, then prints I² = 87% — a figure that means they disagreed considerably, and which by the paper's own stated method should have called for a different statistical model. The finding stands; it's simply less settled than the write-up suggests. The full text is open access at the link below for anyone who wants to check the arithmetic themselves.
Two findings that couldn't be pooled
Two outcomes appear in the paper without statistics attached, and both are worth knowing precisely because they're incomplete.
Depression. A single trial measured depression using the Hamilton Depression Rating Scale. Scores in the HBOT group fell significantly compared with controls. With only one trial reporting it, no meta-analysis was possible.
Non-motor symptoms overall. Two trials used the Non-Motor Symptom Questionnaire, a broad instrument spanning everything from constipation to anxiety to fatigue. Both found lower scores in the HBOT groups. Two trials aren't enough to pool.
Both point the same way as the pooled results, which is encouraging. Neither carries statistical weight on its own — they're the kind of early signal that tells researchers where to look next.
Side effects: what little was reported
Three trials reported on adverse events, two of them with full statistical comparison. Neither found any significant difference between the HBOT and control groups.
The more important number is the other one: ten of the thirteen trials didn't report on adverse events at all. That isn't evidence of safety, it's an absence of data — and it sits awkwardly next to the toxicity literature described earlier. A trial that doesn't record harms cannot tell you there weren't any.
⚖️ Reading these results carefully
First, how good were these thirteen trials?
This is the caveat that frames everything above, and the paper is admirably direct about it.
The authors scored every included trial on the Jadad scale, the standard instrument for grading trial quality. The verdict: the overall quality of the included literature was low.
The specifics matter more than the score. All thirteen trials described themselves as randomized. Six explained how — they used a random number table. Seven gave no description of their randomisation method at all. In a randomized controlled trial, how patients were assigned to groups is the mechanism that makes the whole comparison meaningful. Seven trials simply assert it happened.
It's also worth looking at what sits underneath the headline number. The 91% versus 76% comparison rests on four trials. The largest of them was published in 1995, and it is the one trial in the whole analysis that never specified how its patients were diagnosed with Parkinson's — the other twelve used China's national diagnostic criteria. Another of the four isn't a journal paper at all but a two-page conference abstract covering ten patients per group. The finding is real and the four trials genuinely agree with each other. They are simply not four equally solid trials.
Only one trial out of thirteen reported how many participants dropped out and why. In the other twelve, patients who left are invisible — and people who leave a trial are rarely a random sample of the people who stay.
None of this means the findings are wrong. It means the evidence sits several rungs below what a Cochrane-grade review would call firm, and everything above should be read as a consistent signal from a thin literature rather than as an established effect.
Every trial came from one country
All 13 trials were conducted in China, published between 1995 and 2022, mostly in Chinese-language journals. The researchers applied no geographic restriction — trials from elsewhere simply don't exist.
This is a recurring pattern in HBOT research. It matters because replication across different healthcare systems, patient populations, and research cultures is how findings get stress-tested. That hasn't happened for Parkinson's.
Blinding — and an interesting pattern in the data
None of the trials were blinded. The authors state this plainly, noting that genuine double-blinding is nearly impossible in hyperbaric research, since patients know whether they're in a pressurised chamber.
But there's a pattern in these results worth noticing. The findings with the tightest agreement between trials — overall treatment response, daytime sleepiness, cognitive testing, disease staging — are all measures with an objective or clinician-scored component. The two findings with the shakiest statistics are motor scoring (I² = 77%) and sleep quality (I² = 87%), the latter a pure self-report questionnaire.
If the entire effect were driven by patient expectation, you'd expect the opposite: the subjective measures showing the cleanest, most consistent improvements. That doesn't prove the effect is real. It's a data pattern that runs against the simplest sceptical explanation.
What the dose question would take to settle
There is one pattern worth noting here, with a caveat attached. The four trials behind the treatment-response finding — the one with perfect agreement between studies — all used pressures of 2.3 to 2.5 ATA. The nine trials measuring motor function, where agreement was poor, mostly used 2.0 ATA.
It's tempting to read that as higher pressure producing more consistent results. But there's a simpler explanation: treatment response is a yes-or-no judgment — did the patient improve by 11% or more? — and binary measures naturally scatter less than continuous scales like UPDRS. The clean agreement may be an artefact of how the outcome was measured rather than anything about pressure at all.
There's a small irony in the paper's own reference list. Among the studies the authors cite while building their case is a 2018 mouse experiment in which mild hyperbaric oxygen slowed the loss of dopamine neurons in the substantia nigra. The lower end of the range has preclinical support that the human trials collected here never followed up on.
The authors put the limitation plainly: "there is no consensus on the optimal dosage and duration of HBO therapy, which hampers its application." Until trials record what they actually did — pressure, session length, frequency, total sessions — and until someone tests the lower end of the range against the higher, the most important question in this literature stays open.
📌 Takeaway for the community
- A meta-analysis of 13 randomized trials with 958 Parkinson's patients found that adding HBOT to standard medication produced meaningful improvement in 91% of patients, versus 76% on medication alone — with all four contributing trials in complete agreement
- Daytime sleepiness dropped 3.3 points on a 24-point scale, enough to move a typical patient back inside the normal range, and cognitive function improved with a moderate-to-large effect size — both with zero disagreement between studies
- Motor function improved by 2.96 points, which sits at the lower edge of what patients actually notice, and the trials disagreed on the size of the effect
- Disease staging shifted by an average of 0.14 — roughly one patient in seven moving down a full stage — though Hoehn-Yahr is scored from observed symptoms, so this is not evidence of disease modification
- The trials were of low methodological quality: seven of thirteen never described how they randomised patients, and only one reported dropouts
- Every reported pressure was 2.0 ATA or above, two trials didn't report pressure at all, and no trial reported session length or frequency — so the dose these patients received is unknown, and nobody tested whether lower pressures would have done better
- These studies were conducted in clinical hospital settings under medical supervision — results cannot be directly applied to other settings or devices
Source: https://karger.com/dem/article/54/3/187/916434/
Bu S, Liu W, Sheng X, Jin L, Zhao Q. Hyperbaric Oxygen Therapy Improves Motor Symptoms, Sleep, and Cognitive Dysfunctions in Parkinson's Disease. Dement Geriatr Cogn Disord. 2025;54(3):187-200. doi: 10.1159/000542619.
Educational disclaimer
This content summarizes findings from published medical research for educational purposes only.
The hyperbaric chambers sold on this website are non-medical wellness devices and are not intended to diagnose, treat, cure, or prevent any disease.
The studies discussed here were conducted in clinical medical settings using medical-grade interventions. The inclusion of research summaries does not imply that similar outcomes can be achieved using non-medical wellness devices.
