Long COVID Biology — Why Symptoms Persist

When SARS-CoV-2 emerged in 2020, public attention centered on the acute illness: respiratory failure, intensive care, and death. Within months a second pattern appeared. Some people, including many whose first infection was mild, did not recover on schedule. They stayed ill with fatigue, cognitive trouble, and breathlessness that standard tests often failed to explain. Long COVID is the name for that pattern. This article explains what researchers mean by the term, how common it is, which biological mechanisms are under study, and what the evidence does and does not show.

Educational content, not medical advice. Talk to your doctor before changing any medication, treatment, or routine. Information here reflects published research, but every person's situation differs.

What counts as long COVID

There is no blood test for long COVID, so the condition is defined by symptoms and time. Three definitions are in wide use, and they differ in ways that matter when you compare studies.

Source	Year	Core wording	Time threshold
WHO Delphi consensus	2021	History of probable or confirmed SARS-CoV-2 infection, symptoms that last at least 2 months and cannot be explained by an alternative diagnosis	Usually appears about 3 months after COVID-19 onset
RECOVER (NIH) research definition	2023	A score built from 12 symptoms reported by participants	Assessed at 6 months
National Academies (NASEM)	2024	An infection-associated chronic condition that continues, relapses and remits, or progresses	At least 3 months

The World Health Organization's definition came out of a Delphi process, a structured method in which an expert panel votes through successive rounds until it converges. WHO says it was developed by patients, researchers, and others representing all WHO regions. The National Academies report of 2024 framed long COVID as one example of an infection-associated chronic condition, a category that includes conditions that follow other infections. It does not require laboratory confirmation of the original infection, and the three months need not run continuously from the acute illness.

The 2023 RECOVER study, published in JAMA by Thaweethai and colleagues, took a different route. It analyzed 9,764 adults, 89 percent of whom had been infected, and used statistics to find which symptoms best separated infected from uninfected participants. Twelve symptoms formed the score: post-exertional malaise, fatigue, brain fog, dizziness, gastrointestinal symptoms, palpitations, changes in sexual desire or capacity, loss of smell or taste, thirst, chronic cough, chest pain, and abnormal movements.

Think of it like diagnosing a car problem from a driver's description alone, with no diagnostic port to plug into. A shared definition lets everyone describe the same rattle, but it does not say which part is broken.

How common is it

Prevalence depends heavily on the definition, the survey method, and the variant that was circulating. The figures below come from different studies and cannot be added together or compared directly.

Measure	Result	Source
US adults reporting ever having long COVID, 2022	6.9 percent (about 18 million people)	CDC National Health Interview Survey
US adults with long COVID symptoms at time of survey, 2022	3.4 percent	CDC National Health Interview Survey
US adults, 2023 survey	8.4 percent ever; 3.6 percent currently; 2.3 percent currently with activity-limiting symptoms	CDC research letter in JAMA Network Open
RECOVER participants first infected on or after December 1, 2021	224 of 2,231 (10 percent) met the PASC score at 6 months	Thaweethai et al., JAMA, 2023

In the CDC's 2022 survey, women reported long COVID more often than men (8.5 percent versus 5.2 percent), and adults aged 35 to 49 reported the highest rates (8.9 percent). The 2023 figures added a question about activity limitation, so they are not a like-for-like comparison with 2022.

Two factors shifted risk over time. The first is variant. A study using the UK COVID Symptom Study app compared 56,003 adults infected during the Omicron wave with 41,361 infected during the Delta wave. Long COVID prevalence was 4.4 percent in the Omicron group and 10.8 percent in the Delta group, and the odds of long COVID were 20 to 50 percent lower with Omicron depending on age and time since vaccination. That is a self-reported, app-based study, and the authors noted that the absolute number of people affected also depends on how many were infected.

The second is vaccination. A meta-analysis in the Journal of Infection pooled 25 studies through February 2024 with more than 14 million participants. Two doses before infection were associated with 24 percent lower odds of long COVID (odds ratio 0.76), and one dose after infection with 15 percent lower odds (odds ratio 0.85). All the included studies were observational, so the result shows association, not proof of cause. People who get vaccinated may differ from those who do not in ways that also affect long COVID risk.

What long COVID looks like

Dozens of symptoms have been linked to the condition (the RECOVER analysis found 37 that were more common in infected participants), but a handful dominate. The ones with the strongest research attention are:

• Post-exertional malaise (PEM). A disproportionate worsening of symptoms after physical, cognitive, or emotional effort, often delayed by a day or more. It differs from ordinary tiredness because the crash is out of proportion to the effort and recovery takes days.
• Fatigue. Persistent exhaustion that sleep does not relieve.
• Brain fog. Slowed thinking, memory lapses, and trouble finding words.
• Breathlessness and chest symptoms, sometimes without measurable lung damage.
• Autonomic dysfunction. The autonomic nervous system controls heart rate, blood pressure, and digestion without conscious effort. In postural orthostatic tachycardia syndrome (POTS), the heart rate rises by 30 beats per minute or more within 10 minutes of standing, without a drop in blood pressure.
• Sleep disruption, headaches, and pain.

Because patients differ so much, many researchers believe long COVID is several overlapping conditions under one label. A person with POTS and PEM may have a different dominant mechanism from a person whose main problem is smell loss. The RECOVER-VITAL trial, described below, recruited participants in three groups by dominant symptom (cognitive dysfunction, autonomic dysfunction, or exercise intolerance), for this reason.

Six proposed mechanisms

The research literature points to several mechanisms that could operate alone or in combination. The sections below describe each, the best evidence, and its limits.

1. Persistent virus or viral fragments

The idea: After acute infection, some virus or viral material stays in tissue and keeps the immune system activated. The closest analogy is embers left in a fireplace. The visible fire is out, but the embers can reignite or keep the room warm.

The evidence:

• A 2022 Nature autopsy study by Stein and colleagues examined 44 people who died with COVID-19. Viral RNA was found throughout the body, including regions across the brain, and was detectable up to 230 days after symptom onset in some patients. Little damage to brain tissue was seen. The study did not connect persistence to long COVID symptoms.
• A 2022 study in Gastroenterology by Zollner and colleagues took gut biopsies from 46 people with inflammatory bowel disease who had a prior infection. About seven months after mild acute infection, viral RNA was found in the gut lining of 32 patients and viral nucleocapsid protein in 24. The team could not culture live virus from that tissue, and the study was small and limited to a specific patient group.

The test: If persistent virus drives symptoms, antivirals should help. The NIH's RECOVER-VITAL trial tested this. It randomized 959 adults with long COVID at 69 US sites to 25 days of nirmatrelvir-ritonavir (Paxlovid), 15 days of Paxlovid followed by 10 days of ritonavir alone, or 25 days of ritonavir alone, and followed them for six months. Results published in The Lancet Infectious Diseases in August 2026 reported that Paxlovid did not improve long COVID symptoms compared with placebo, with no serious safety concerns from the longer course. Co-principal investigator Lindsey Baden said that treating viral persistence with this antiviral did not show evidence of clinical benefit.

What it means: The trial tested one drug at one dose in people with long COVID. It does not rule out a persistence role for some patients or a different antiviral approach, but it weakens the simplest version of the hypothesis.

2. Immune dysregulation and autoimmunity

The idea: The immune system fails to switch off, or it turns on the body's own tissue. A smoke alarm that keeps ringing after the fire has been put out is a fair picture.

The evidence:

• A 2023 Nature paper by Klein and colleagues profiled 275 people with and without long COVID. Those with long COVID showed differences in circulating myeloid and lymphocyte populations, stronger antibody responses to SARS-CoV-2, higher antibody responses to other viruses including Epstein-Barr virus, and lower cortisol levels. Cortisol is a stress hormone released through the hypothalamic-pituitary-adrenal axis, which we cover in our guide to cortisol and the stress response.
• A 2022 study in Cell by Su and colleagues followed 309 people from COVID-19 diagnosis through two to three months of recovery. Four factors measured at diagnosis anticipated later symptoms: type 2 diabetes, circulating SARS-CoV-2 RNA, Epstein-Barr virus viremia, and certain autoantibodies. Autoantibodies are antibodies that bind the body's own proteins.

For background on how antibodies and immune cells normally work, see our guides to how the immune system fights viruses and antibody structure.

Limits: These are association studies. Finding an autoantibody in a patient does not show that it causes a symptom, since infections can produce autoantibodies that do nothing. The immune differences could also be a consequence of ongoing illness rather than its cause.

3. Reactivation of latent viruses, especially Epstein-Barr virus

The idea: Many people carry herpesviruses in a dormant state. Infection or immune disturbance could wake them. Picture a sleeping dog in the house: a noisy visitor, in this case SARS-CoV-2, may be enough to rouse it.

Epstein-Barr virus (EBV) is the most studied candidate because it infects most adults and has links to other post-infectious conditions, as described in our explainer on Epstein-Barr virus. Our overview of the herpesvirus family covers the other members that establish lifelong latency.

The evidence: The Su study found EBV viremia at diagnosis was one of four predictors of later symptoms, and the Klein study found stronger antibody responses against EBV in long COVID. Antibody levels can rise because of reactivation, but they can also reflect an immune system that is generally overactive, so they are an indirect signal.

Limits: Reactivation is a hypothesis with supporting correlations. No trial in the sources reviewed here has shown that suppressing EBV improves outcomes.

4. Microclots and blood vessel dysfunction

The idea: Tiny, abnormal clots might block the smallest blood vessels, reducing oxygen delivery to tissue. Imagine sand in the gears of a watch: the large parts look fine, but fine movement fails. Our guide to how blood works explains normal clotting.

The evidence: Resia Pretorius and colleagues reported fluorescent "amyloid microclots" in blood from people with long COVID, and the group has continued to publish on the topic.

Limits: This is the most contested mechanism. The findings come largely from one research group, the quantification methods are not standardized, and microclots have not been established as a cause of long COVID in patients. Treat the hypothesis as open.

5. Gut microbiome disruption

The idea: SARS-CoV-2 enters cells through a receptor called ACE2, which is abundant in the gut lining. Infection might change the bacterial community in the intestine, which in turn affects immune signaling. For more on that community, see our explainer on the gut microbiome.

The evidence: A 2022 study in Gut by Liu and colleagues followed 106 patients from hospital admission to six months, alongside 68 uninfected controls, and analyzed 258 stool samples. At six months, 76 percent of patients reported persistent symptoms. Those with symptoms had higher levels of Ruminococcus gnavus and Bacteroides vulgatus and lower levels of Faecalibacterium prausnitzii, a butyrate-producing bacterium. Patients without persistent symptoms had microbiomes close to those of controls.

Limits: The authors describe the result as an association. The microbiome may be changed by hospitalization, antibiotics, and diet, all of which accompany severe COVID, and the study cannot say which came first.

6. Brain changes and neuroinflammation

The idea: Brain fog may reflect structural or functional changes in the brain. A 2022 Nature study by Douaud and colleagues used a rare opportunity: the UK Biobank had scanned participants before the pandemic, so researchers could compare brains before and after infection.

The evidence: The study included 785 people aged 51 to 81, scanned twice about 3.2 years apart. Of these, 401 tested positive for SARS-CoV-2 between scans and 384 did not. The infected group showed greater tissue loss in regions linked to smell and memory, including the orbitofrontal cortex, the parahippocampal gyrus, and the insula, with average regional differences of 0.2 to 2.0 percent. They also showed more cognitive decline. Infections occurred between March 2020 and April 2021. The differences persisted when the 15 hospitalized patients were excluded.

Limits: Most infections were mild and occurred in 2020 and early 2021, when the Alpha and Beta variants dominated. Omicron was not studied. The authors could not say whether the changes are reversible, and the study was not designed to link them to long COVID symptoms. Our guide to how the aging brain changes gives a baseline for what normal brain volume loss looks like.

Comparing the mechanisms

Mechanism	Strongest evidence	Main weakness	Status
Viral persistence	Autopsy RNA up to 230 days (Stein); gut antigen (Zollner)	Persistence not tied to symptoms; RECOVER-VITAL found no benefit from Paxlovid	Plausible for some patients; simplest version weakened
Immune dysregulation	Immune profiling (Klein); early predictors (Su)	Association, not causation	Well supported as a feature
EBV reactivation	EBV viremia and antibody signals	Indirect measures; no treatment trial result	Hypothesis with correlations
Microclots	Lab detection of fibrin-containing particles	Non-standardized methods; mostly one research group	Contested
Gut microbiome	Altered bacteria at six months (Liu)	Confounded by hospitalization and drugs	Association
Brain changes	Before-and-after MRI (Douaud)	Mostly mild early-variant cases; link to symptoms untested	Measurable differences, cause unknown

The overlap with ME/CFS

Myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) is a long-term illness defined by profound fatigue, post-exertional malaise, unrefreshing sleep, and cognitive difficulty. It has long followed infections. Post-exertional malaise in long COVID resembles the ME/CFS pattern closely, and many researchers see the two as overlapping, which is part of why the National Academies put them under the shared heading of infection-associated chronic conditions.

The overlap has a practical consequence for rehabilitation. Programs that increase exercise on a fixed schedule were once common for chronic fatigue, but they are questioned for patients with post-exertional malaise, who can be worsened by exertion. Pacing, which means keeping activity within an individual limit to avoid crashes, is the approach most often discussed in its place. The evidence for pacing is still limited, and decisions about activity belong with a clinician who knows the individual.

What treatments have been tested

No therapy has been established as effective for long COVID. The table lists what the trials described above have shown, and it is not a recommendation.

Approach	What was tested	Result
Extended Paxlovid	RECOVER-VITAL, 959 adults, 25 days	No improvement over placebo (Lancet Infectious Diseases, August 2026)
Vaccination before infection	Observational meta-analysis, 25 studies	Lower odds of long COVID (odds ratio 0.76), association only

Many other approaches, from anticoagulants to repurposed drugs, are in earlier trials. A result that comes from an open-label study with no placebo group is weak evidence in a condition where symptoms come and go.

Common misconceptions

"Long COVID only follows severe COVID." Severity of the acute illness raises risk, but long COVID occurs after mild infections too. The RECOVER cohort and the UK Biobank imaging work both included many people with mild disease.

"It is a psychological condition." Studies have documented measurable immune, gut, and brain differences. That does not mean mood symptoms are absent or unimportant. Anxiety and low mood are common in chronic illness, and our explainer on the neuroscience of anxiety shows how stress and body physiology interact. But reducing a condition with documented biological findings to a psychological one is not supported by the evidence.

"Microclots are the proven cause." They are one proposed mechanism with disputed methods. Claims that a clot-targeting treatment works should be read against the conflicting findings.

"One test will soon diagnose it." Researchers are searching for biomarkers, but none has been validated for routine diagnosis. A single marker may never suffice if long COVID comprises several subtypes.

"If it were persistence, antivirals would have worked." The trial result is informative but narrow. One drug at one dose in a mixed population is not a test of every way a reservoir could matter.

Why long COVID is hard to study

Several features make clean answers slow to arrive.

• No gold-standard test. Without a biomarker, definitions rely on symptoms, which differ between studies.
• Heterogeneity. If several mechanisms exist, a drug that helps one subgroup may look ineffective in a mixed trial.
• Self-report. Most large surveys depend on people describing their own symptoms, which introduces variation in how questions are asked.
• Moving target. Variants, vaccination, and prior infections have all changed since 2020, so results from early cohorts may not carry over.
• Fluctuating symptoms. Placebo responses and natural ups and downs can mask or mimic a drug effect, which is why placebo-controlled designs like RECOVER-VITAL matter.

Sleep, hormones, and the nervous system all interact in ways that complicate causal claims. Our guide to the science of sleep shows how disrupted rest can amplify other symptoms regardless of the original trigger.

Frequently asked questions

Q: Can long COVID occur after a mild infection? A: Yes. Long COVID is not limited to people who were hospitalized. The UK Biobank imaging study found brain differences in a group that was mostly mild, and the RECOVER cohort included people with a range of acute severity. Acute severity affects risk, but it does not predict who will develop long COVID.

Q: Does vaccination reduce the risk? A: Observational data point that way. A pooled analysis of 25 studies found 24 percent lower odds of long COVID with two doses before infection. Because all the studies were observational, they cannot prove vaccination caused the reduction, and protection is not complete. The authors called for more work on booster doses in the Omicron era.

Q: Is long COVID a psychological condition? A: The evidence does not support that view. Researchers have documented immune differences, gut microbiome changes, and brain imaging differences in groups with long COVID. Mental health support can still help people cope with any chronic illness, which is a separate point.

Q: How long does long COVID last? A: It varies. The definitions in use require symptoms for at least two or three months, and the National Academies framing includes courses that are continuous, relapsing and remitting, or progressive. Large long-term outcome data are still accumulating, and outcomes in early-pandemic cohorts may not match those of people infected more recently.

Q: What is POTS, and how is it related? A: Postural orthostatic tachycardia syndrome is a disorder of the autonomic nervous system in which heart rate rises by 30 beats per minute or more within 10 minutes of standing, without a fall in blood pressure. It occurs in a subset of people with long COVID. Proposed causes include autoantibodies, nerve involvement, and blood volume changes, and none is confirmed.

Q: How is long COVID diagnosed? A: By a clinician, using a patient's history and by ruling out other explanations. The WHO and National Academies definitions both rest on symptoms and duration. No blood test or scan can confirm it, though researchers are testing candidate biomarkers.

Q: Why did the Paxlovid trial matter if it failed? A: A negative result from a well-designed, placebo-controlled trial narrows the field. RECOVER-VITAL showed that the simplest form of the persistence hypothesis, where a few more weeks of antiviral treatment clears the problem, did not hold in this setting. That redirects effort toward other mechanisms and toward trials that match treatments to patient subgroups.

The takeaway

Long COVID is best understood as a syndrome with several plausible biological routes, not a single disease with a single cause. Evidence is strongest that the immune system is altered and that some viral material can linger in tissue, but neither has been shown to explain symptoms in everyone. The largest antiviral trial so far has not delivered a treatment, which shifts the bet toward subtypes and combinations. For now, the most reliable facts are modest ones: it is common enough to be measured in national surveys, it follows mild infections, vaccination before infection is associated with lower risk, and its biology is real enough to study with the same rigor as any other chronic disease.

Sources

1. World Health Organization. A clinical case definition of post COVID-19 condition by a Delphi consensus, October 6, 2021.
2. National Academies of Sciences, Engineering, and Medicine. A Long COVID Definition: A Chronic, Systemic Disease State with Profound Consequences, 2024.
3. Thaweethai T, et al. Development of a definition of postacute sequelae of SARS-CoV-2 infection. JAMA, 2023. RECOVER summary.
4. Klein J, et al. Distinguishing features of long COVID identified through immune profiling. Nature, 2023.
5. Su Y, et al. Multiple early factors anticipate post-acute COVID-19 sequelae. Cell, 2022.
6. Stein SR, et al. SARS-CoV-2 infection and persistence in the human body and brain at autopsy. Nature, 2022. CIDRAP summary.
7. Douaud G, et al. SARS-CoV-2 is associated with changes in brain structure in UK Biobank. Nature, 2022. Commentary letter.
8. Liu Q, et al. Gut microbiota dynamics in a prospective cohort of patients with post-acute COVID-19 syndrome. Gut, 2022. Gut full text.
9. RECOVER-VITAL trial results announcement, August 2026.
10. Antonelli M, et al. Risk of long COVID associated with delta versus omicron variants of SARS-CoV-2. The Lancet, 2022. King's College London summary.
11. CIDRAP: analysis of 25 studies on vaccination and long COVID, Journal of Infection.
12. CIDRAP: about 8% of US adults have ever had long COVID, reporting CDC survey data.

---

Read next:

• ADHD: The Neuroscience Behind the Diagnosis
• The Biology of Aging and Longevity Science
• Alzheimer's Disease: The Biology Behind the Decline