This post is a companion to Post #34, which laid out the exercise paradox: Long COVID patients are told to exercise, but the data shows exercise can make a subset measurably worse. That post was about the ~40% with objectively confirmable post-exertional malaise.
This post is about the other ~60%. The ones who don't crash. The ones who could benefit from structured rehabilitation — if anyone had properly studied it.
The Evidence Base Is One Trial
Five years into Long COVID research, with $1.15 billion in RECOVER funding alone, we have exactly one randomized controlled trial that did what should be obvious: test exercise rehabilitation in Long COVID patients after first excluding those for whom exercise is contraindicated.
Scharitzer et al. (2025) enrolled 42 patients in Vienna. They explicitly excluded patients with post-exertional malaise. Then they ran a 12-week supervised rehabilitation program — aerobic training, resistance work, respiratory exercises — and compared it to usual care.
The effect size on VO₂peak was 1.75. In clinical trial terms, that's enormous. For context, anything above 0.8 is considered "large." Fatigue improved. Quality of life improved. No adverse events related to the intervention.
Forty-two people. One trial. That's the entire rigorous evidence base for exercise in the non-PEM Long COVID population.
Why This Isn't Deconditioning
The most common dismissal of Long COVID exercise intolerance is deconditioning — patients are inactive, they lose fitness, they feel worse, they become more inactive. The implication is that the problem is behavioral. Just move more.
Two papers from the same research group at Workwell Foundation dismantle this.
Thomas et al. (Experimental Physiology, 2026) put 68 Long COVID patients through a two-day submaximal cardiopulmonary exercise test. On Day 2, their ventilatory threshold (VT1) oxygen consumption dropped significantly (P=0.003) and oxygen pulse dropped (P=0.002). Deconditioned people don't get worse on Day 2. They get the same result. This is physiological deterioration with repeat exertion — the metabolic signature of an oxygen delivery system under stress.
Then came Thomas et al. (Physiological Reports, April 2026): same group, same participants, this time with near-infrared spectroscopy (NIRS) sensors on the gastrocnemius muscle. In controls, tissue oxygen saturation stayed elevated above resting levels for 5 minutes during submaximal exercise. In Long COVID patients, it lasted 2 minutes on Day 1 and collapsed to 1 minute on Day 2. Ventilatory efficiency was 17% worse. Respiratory exchange ratio was elevated. The oxygen wasn't getting to the muscles.
The critical nuance: this NIRS study excluded participants with moderate-to-severe post-exertional malaise using the DePaul screening tool. These are Group B patients — the ones who don't crash severely. And they still have measurable peripheral oxygen delivery dysfunction. The deficit is real, it's physiological, it's not deconditioning — and the Vienna trial shows it's trainable.
This is the argument the field hasn't assembled from its own data. The Thomas papers prove the problem is real. The Scharitzer trial proves it responds to structured intervention. The gap between these findings — real deficit, real response — is where clinical rehabilitation should live. But almost nobody is there, because the field hasn't distinguished the patients who crash from the patients who could improve.
The VT1 Boundary
Ventilatory threshold 1 — VT1 — is the exercise intensity at which breathing shifts from comfortable to labored. Below it, you're burning mostly fat through aerobic metabolism. Above it, anaerobic contribution rises, lactate accumulates, and in Long COVID patients with impaired oxygen delivery, the system fails.
Charlton et al. (BJSM, March 2026) argued that VT1 should be the hard ceiling for Long COVID exercise prescription — not arbitrary heart rate zones, not perceived exertion, not the standard fitness formulas. The physiological boundary. Stay below where oxygen delivery fails, and you're training the system. Cross it, and you're provoking the exact metabolic crisis that triggers post-exertional symptoms.
The Thomas NIRS data gives this clinical reasoning a mechanistic anchor. VT1 isn't an abstract threshold — it's the point at which peripheral tissue oxygenation collapses. The NIRS sensors can see it happening in real time at the muscle.
Heart rate pacing approximates this boundary. Clague-Baker et al. showed 89% adherence at 8 weeks with a simple HR-monitored pacing protocol (n=47). Not a clinical trial — a feasibility study. But it demonstrates that patients can learn to identify and respect the boundary without exercise lab equipment.
What the Meta-Analyses Show (and Hide)
The broader literature is not empty. Silveira et al. (2026) found 15 studies with 803 participants showing that supervised outpatient pulmonary rehabilitation improved functional exercise capacity (6-minute walk distance +53.72 meters) and quality of life. Fatigue trended toward improvement across heterogeneous measures. A separate meta-analysis found an effect size of 0.89 for fatigue reduction — solidly in the "large" range.
But these numbers hide the critical problem: almost none of these studies screened for or excluded PEM. A Sick Times investigation found that fewer than 20% of Long COVID exercise trials even mention post-exertional malaise. The BMC three-tiered protocol review noted that "none of 46 studies specifically reported on PEM."
This means the meta-analytic evidence is contaminated. Patients who would deteriorate with exercise are mixed with patients who would improve. The aggregate effect is positive because the improvers outnumber the crashers — but the aggregate tells you nothing about which patient is which. If you're a clinician reading the meta-analysis, you don't know whether your patient is a Vienna responder or a Thomas Day-2 decliner.
The AHA Scientific Statement (Circulation, June 2025) recommended starting rehabilitation recumbent, progressing gradually, monitoring for PEM. The ME Association immediately challenged it, arguing that the Statement's deconditioning framing ignores the biological evidence for exercise intolerance. Both sides are partially right. The AHA's recommendations are reasonable for Group B patients. They're dangerous for Group A patients. The Statement doesn't make the distinction clearly enough.
RECOVER-ENERGIZE Will Answer This
The RECOVER-ENERGIZE trial is the first properly designed large-scale study. It enrolls 660 participants and does what should have been done in 2021: separates patients with PEM from patients without PEM and tests different interventions in each group. Cardiopulmonary rehabilitation for those without significant PEM. Structured activity pacing for those with PEM.
Results are expected late 2026. When they arrive, we'll have an answer — not from 42 patients in Vienna, but from a properly powered, PEM-stratified, multicenter trial. Until then, the Vienna trial stands alone as the only rigorous positive evidence, and the Thomas NIRS data stands as the best explanation of why the intervention works when it does.
What This Means Right Now
If you have Long COVID and you don't experience post-exertional malaise — no crashes after activity, no 24-48 hour delayed worsening — the evidence, thin as it is, supports structured rehabilitation. Stay below VT1. Start recumbent if necessary. Use heart rate monitoring. Progress slowly. The effect size is large and the adverse event profile in properly screened patients is clean.
If you're unsure whether you have PEM, get a proper assessment. The DePaul Symptom Questionnaire screens for it. A two-day cardiopulmonary exercise test with VT1 measurement can confirm it physiologically. This matters — misclassification isn't a statistical inconvenience, it's the difference between rehabilitation and relapse.
And if you do have PEM, this post isn't for you. Post #34 is. Pacing, not pushing. The Thomas Day-2 data shows what happens when the oxygen delivery system is pushed past its limit, and it gets worse, not better, with repetition.
The scandal isn't that exercise doesn't work for Long COVID. For a majority of patients, it probably does. The scandal is that after five years and billions in research funding, we have one properly designed trial with 42 participants to prove it. The Vienna trial's effect size of 1.75 shouldn't be reassuring — it should be an indictment of a research infrastructure that spent five years not asking the obvious question.