Long COVID Guideline: Why Exercise Must Wait for PEM Screening

The most consequential Long COVID recommendation may sound modest: do not prescribe exercise rehabilitation until you have ruled out post-exertional malaise (PEM). For people with PEM, even modest activity can trigger a disproportionate symptom flare. A fitness plan built around steadily adding workload can therefore make the wrong problem worse.

That warning comes from a 2026 European Respiratory Journal clinical practice guideline led by Bin Cao and developed by a 60-expert panel from 10 countries. The panel used GRADE, a framework that rates both the certainty of evidence and the strength of recommendations. It addressed eight formal clinical questions and issued 10 recommendations. None rests on high-certainty evidence. That is the central fact, and it should temper both therapeutic optimism and therapeutic nihilism.

A guideline built to show uncertainty

The panel conditionally recommends pre-infection COVID-19 vaccination, multi-species probiotics, cognitive behavioral therapy (CBT) for fatigue, and personalized rehabilitation after PEM has been excluded. It conditionally recommends against nirmatrelvir-ritonavir (Paxlovid) for established Long COVID and against glucocorticoids for persistent respiratory symptoms and olfactory disorders.

"Conditional" matters. It does not mean an intervention failed or succeeded beyond dispute. It means the expected benefits, harms, patient preferences, and limits of the evidence do not support a blanket recommendation. Across the guideline, certainty ranged from very low to moderate. Long COVID care has accumulated enough evidence to reject a few easy assumptions, but not enough to offer a reliable universal treatment.

Rehabilitation depends on a safety question first

Personalized rehabilitation may help people who do not have PEM, according to the guideline. But the panel puts the safety condition before the exercise prescription: exclude PEM first.

PEM is more than ordinary post-workout soreness or feeling tired after a busy day. The guideline describes disproportionate symptom worsening after exertion; the clinical concept document places this flare in the 12-to-48-hour period after activity. Fatigue, cognitive symptoms, breathlessness, pain, and other symptoms can rise after a delay, which makes a standard gym-style progress log a poor safety monitor.

The guideline does not specify a validated screening instrument for operationalizing that exclusion. That leaves clinicians with an important principle but incomplete instructions. Still, the practical implication is clear: someone whose symptoms reliably worsen after physical, cognitive, or emotional exertion needs assessment and pacing-oriented management, not an automatic graded-exercise program.

This distinction also explains why wearable data deserve interest without becoming a treatment plan. In a 2025 secondary analysis of the STOP-PASC randomized trial, adults with symptoms lasting at least three months after SARS-CoV-2 infection used digital measures including resting heart rate, step count, and sleep architecture. These measures offered continuous, objective proxies for fatigue, PEM, and shortness of breath. They can document physiological patterns that questionnaires may miss. They do not, by themselves, prove a mechanism or tell a patient how much activity is safe.

Paxlovid does not have evidence for treating established Long COVID

The guideline conditionally recommends against nirmatrelvir-ritonavir for Long COVID treatment. This is easy to misread. It does not say Paxlovid lacks value for acute COVID-19 in people who meet treatment criteria. It says evidence does not support using it to treat established Long COVID.

STOP-PASC is central to that conclusion, yet the published digital-biomarker analysis does not report the trial's primary symptom-efficacy comparison between nirmatrelvir-ritonavir and placebo. It reports wearable findings, not the drug's primary clinical effect estimate, confidence interval, or p-value. Readers should therefore treat the guideline's conditional negative recommendation as a careful clinical judgment, while recognizing that the accessible publication does not let them independently inspect the central efficacy result.

The same restraint applies to glucocorticoids. The panel recommends against steroids for persistent respiratory symptoms and smell disorders, reversing the instinct to extend acute-COVID treatment into the post-acute phase. The available evidence does not identify a purpose-designed Long COVID steroid trial or establish whether this recommendation reflects demonstrated harm, lack of benefit, or both. "Steroids helped in acute severe COVID" is not evidence that they help months later.

Two conditional treatments, neither ready for self-prescribing

The guideline gives conditional positive recommendations to multi-species probiotics and CBT for Long COVID fatigue. CBT is directed at the fatigue domain; it should not be presented as a cure for every Long COVID symptom or as proof that symptoms are psychological. The guideline supports it as one option for people whose fatigue is the main disabling problem.

The probiotic recommendation is even less actionable for consumers. The recommendation does not provide strain-level composition or dosing in the published material described here, and no independently available randomized trial with a defined multi-species formula establishes which product to buy. A shelf full of probiotic labels is not a clinical trial. Until strains, dose, duration, and patient population are specified, the recommendation is a research signal rather than a shopping list.

Vaccination remains the best-supported prevention strategy

Pre-infection vaccination is the guideline's main Long COVID prevention recommendation. The most directly relevant support comes from Tannous et al.'s 2023 observational registry analysis, which used electronic medical records in a diverse U.S. metropolitan population to evaluate COVID-19 vaccines and anti-SARS-CoV-2 monoclonal antibodies against documented post-acute sequelae. Its direction of effect supports vaccination as a prevention strategy, but its case definition depended on people seeking care and receiving documentation. That can miss cases and introduce socioeconomic confounding.

Other evidence provides context rather than a direct Long COVID answer. The COVE trial's open-label and booster follow-up documented sustained mRNA-1273 immunogenicity and reported 93.2% efficacy against acute COVID-19. Long COVID prevention was not a primary or prespecified secondary endpoint, so acute-infection protection cannot be converted into a quantified Long COVID-prevention effect.

Likewise, a study of 4,090 healthcare workers found detectable anti-SARS-CoV-2 antibodies two years after vaccination; 85.1% received Pfizer/BioNTech vaccine. That durability finding from Szczepanek et al. supports biological plausibility for ongoing protection, not a direct estimate of Long COVID prevention.

Why vaccination evidence cannot be treated as universal

Most of the prevention evidence centers on mRNA vaccines and largely predates Omicron-specific Long COVID studies. There is no Omicron-stratified randomized trial in this evidence base that measures Long COVID prevention. Current estimates for Omicron-era prevention therefore require extrapolation.

Platform and population also matter. A 2025 cross-sectional QazVac study found heterogeneous clinical manifestations and immune-response profiles among people with post-COVID syndrome. Because it was cross-sectional, it cannot establish that the inactivated-virus vaccine caused those differences or quantify prevention. It does show why a simple "all COVID vaccines have the same Long COVID effect" claim outruns the evidence.

Risk may also differ in people less able to mount a vaccine response. A 2025 systematic review in chronic kidney disease reported impaired humoral and cellular responses to SARS-CoV-2 vaccination and elevated Long COVID vulnerability. Vaccination remains sensible prevention, but protection may be attenuated in this high-risk population.

The useful verdict

Long COVID does not yet have a high-certainty treatment. The 2026 guideline offers a more useful map than a list of miracle protocols: vaccinate before infection; consider CBT for fatigue and carefully defined multi-species probiotics with appropriate clinical guidance; and personalize rehabilitation only after assessing for PEM.

The largest unanswered questions are concrete. Researchers need full primary STOP-PASC efficacy results, standardized Long COVID outcomes, validated PEM assessment tools, strain- and dose-specific probiotic trials, and Omicron-era prevention studies across vaccine platforms. Until then, any clinician or influencer claiming certainty has moved ahead of the evidence.