Half of long COVID patients improved within three years, but only 19% in the multisystem group did—an important distinction for future treatment trials.

Long COVID improvement depends heavily on which symptoms dominate. In a study tracking about 1,020 patients, half had meaningfully improved three years after their first positive COVID-19 test. But just 19% of those with symptoms spanning multiple bodily systems reached that threshold, a gap that could shape how researchers design treatment trials.
The findings, published Nov. 3 in Annals of Internal Medicine, draw on the RECOVER Adult Study, a U.S. research project funded by the National Institutes of Health. Researchers assessed symptoms including fatigue, brain fog, shortness of breath and heart palpitations, then followed participants through in-person appointments and symptom surveys every three months.
Improvement depended on the symptom profile
Participants qualified for the analysis if they scored above 10 on the 30-point 2024 LC Research Index, or LCRI, which measures long COVID symptom burden. Researchers counted a participant as improved after their score fell to 10 or below at two consecutive visits.
By that measure, 36% had improved two years after the infection that preceded their illness. The share reached 50% at three years. For 26%, the LCRI fell to zero, indicating that they reported no symptoms at the survey point.
Those figures conceal large differences between symptom groups. In an earlier analysis, RECOVER researchers used an algorithm to sort participants into five subtypes based on predominant symptoms:
- Loss of smell or taste
- Chronic cough
- Brain fog
- Heart palpitations
- Symptoms affecting multiple bodily systems
About 60% of people in the smell or taste group had significantly improved after three years. The brain fog, heart palpitation and multisystem groups were less likely to improve; the rate for the multisystem group was below 20%, at 19%.
Harrison Reeder, a study co-author and biostatistician at Massachusetts General Hospital, avoided calling the participants “recovered.” People with long COVID have told the RECOVER team that symptoms can fluctuate, and the study did not establish whether people who improved later relapsed.
Age and the infecting virus strain also tracked with different outcomes. Men and women improved at similar rates, while adults younger than 46 were more likely to improve than older participants. People infected with Omicron variants, which emerged in 2021 and became dominant, were more likely to improve than those infected with earlier strains.
The multisystem group resembles ME/CFS
The low improvement rate among people with symptoms across multiple systems points to a subgroup that may need focused study. Their symptoms can involve cardiac, nervous and gastrointestinal systems, among others. Dr. Carmen Scheibenbogen, an immunologist at Charité University Medicine Berlin, said their course resembles myalgic encephalomyelitis/chronic fatigue syndrome, or ME/CFS.
“It looks as if at least a major subgroup is suffering from ME-CFS,” Scheibenbogen said, adding that COVID-19 may have triggered the condition. “This is a disease which is triggered by different infections, but in the end, the disease [itself] is very similar,” she said.
ME/CFS, like long COVID, can affect many bodily systems and remains poorly understood despite recent advances in research. The study doesn’t establish that the multisystem group has ME/CFS, but the resemblance is a reason to examine it more closely.
That distinction matters for clinical research. Scheibenbogen and Reeder said future trials should account for long COVID subtypes when evaluating potential treatments. Reeder suggested the multisystem group may have particular room to benefit from an effective intervention, given its lower rate of improvement over time.
The next research step, both said, is to better characterize long COVID through genetic and protein-level studies. The three-year results offer a clear signal for that work: a single overall improvement figure can obscure very different trajectories, and trial design will need to reflect that.

Discussion
Sign in to join the discussion.