Harrison 2024: HBOT for Chronic Stroke — The Replication Problem

The Harrison 2024 trial

Harrison et al. published a 2024 Canadian multi-centre randomised controlled trial of hyperbaric oxygen therapy for chronic stroke recovery. The trial enrolled 136 participants aged 16-36 years, with stroke occurring 6-36 months before enrollment. The HBOT protocol matched the Efrati Lab design: 40 sessions at 2.0 ATA, 90 minutes per session, five days per week, over approximately 8 weeks.

The study concluded that current evidence does not support the routine application of HBOT for chronic stroke recovery.

Why this matters

This trial is important because it was explicitly designed to replicate the earlier Efrati Lab positive findings — Boussi-Gross 2013 and Efrati 2013 (both PLOS ONE) — in a different population and clinical setting. Those earlier trials, both using 2.0 ATA / 40 sessions, had shown cognitive and functional improvements in post-stroke patients, backed by SPECT and perfusion-MRI evidence of restored cerebral metabolism in previously silent regions.

Harrison 2024 is the first major independent attempt to replicate those findings in a multi-centre North American setting. It did not reproduce the positive signal.

Possible explanations for the divergence

Replication failures in medicine are rarely binary refutations. They typically point to population, protocol, or measurement differences that moderate the treatment effect. For Harrison vs. Efrati, several candidates stand out:

• Population age — Harrison enrolled 16-36 year olds. Efrati Lab cohorts have historically been older with more vascular comorbidity. Younger stroke populations may already have superior rehabilitation outcomes with standard care, reducing the room for additional HBOT benefit.
• Stroke etiology mix — the Harrison cohort included more cryptogenic and dissection-related strokes than the Efrati cohorts, which were predominantly thromboembolic. Different mechanisms may respond differently to HBOT.
• Outcome measures — Harrison's primary endpoint selection differed from the Efrati Lab's. Cognitive and functional measures with varying sensitivity can produce divergent results from the same underlying treatment effect.
• Clinical experience — the Efrati Lab operates at a single highly experienced centre. Multi-centre trials introduce heterogeneity in protocol adherence, patient coaching, and in-chamber supervision.
• Rehabilitation co-interventions — the standard-of-care rehabilitation in Canada vs. Israel may differ enough to affect the control arm baseline, compressing the HBOT effect size.

How this fits the broader stroke HBOT evidence

Only 13 clinical trials of HBOT for stroke are registered as of March 2025, making this a small and relatively early evidence base. The published trials span heterogeneous designs, pressures, and session counts. A bibliometric analysis identifies Shai Efrati as the single most-cited author in the field, with 9 publications and an H-index of 35 — but Efrati's work has not been systematically replicated by independent groups until Harrison 2024.

The current state of the evidence can be summarised as:

• Positive single-centre RCTs (Efrati Lab) using 2.0 ATA / 40 sessions in older post-stroke populations
• At least one negative multi-centre RCT (Harrison 2024) in younger patients using the same protocol
• Mechanistic plausibility (angiogenesis, neuroplasticity) remains intact
• UHMS has not approved stroke as an indication

What this means for patients

Chronic post-stroke HBOT remains off-label and cash-pay. It is a reasonable option for patients who have plateaued with standard rehabilitation and who understand the heterogeneity of likely response. Realistic framing:

• Response is likely non-uniform — some patients benefit substantially, others minimally
• HBOT is an adjunct to (not replacement for) continued physical, speech, and occupational therapy
• The Efrati-protocol dose (2.0 ATA / 40 sessions) is the best-supported option
• Younger, less-impaired patients may have the smallest HBOT effect — Harrison 2024 is a boundary condition suggesting this
• Full-course cost is typically $6,000–$18,000

What the field needs

Progress on chronic stroke HBOT will require:

1. Individual-patient-data meta-analysis combining Harrison, Efrati, Boussi-Gross, and smaller trials to identify responder characteristics
2. Biomarker-based patient selection — perfusion-MRI or SPECT patterns that predict likely benefit
3. Standardised core outcome measures across trials to make results directly comparable
4. Longer follow-up — six-month and 12-month durability of any gains, rather than just end-of-treatment

Bottom line

Harrison 2024 is a serious contribution that tempers but does not eliminate the case for chronic post-stroke HBOT. The Efrati Lab evidence remains clinically meaningful; the Harrison result argues against uniform generalisation. For patients, this means setting realistic expectations and choosing centres experienced with the Efrati-protocol dose. For researchers, it is a call for better patient selection and standardised outcomes — not an argument to abandon the indication.

For the full stroke HBOT evidence overview, see our stroke evidence page and the Boussi-Gross and Efrati 2013 entries in our studies database.