Half of people with long COVID show meaningful improvement by 36 months

About half of adults with likely long COVID one year after a first SARS-CoV-2 infection go on to have a meaningful reduction in symptoms within three years, according to a prospective cohort study. The work, led by Harrison T. Reeder of Harvard University, was published online on 3 November in the Annals of Internal Medicine. It also shows that a much smaller share reach complete symptom resolution over the same period.

Study design

The analysis drew on participants in the RECOVER-Adult programme, a multisite study conducted at 83 sites across the United States. It included 1,019 people who still met criteria for likely long COVID 12 months after their first infection. Likely long COVID was defined using the Long COVID Research Index (LCRI), a symptom-based score.

The primary outcome was the time until a participant had two consecutive study visits at which the LCRI fell below 11, the point below which they no longer counted as having likely long COVID. Requiring two visits in a row guards against counting a temporary dip as genuine improvement. The team also tracked three secondary outcomes: the time until the LCRI dropped by at least 50%, the time until quality of life was rated good or better, and the time until the LCRI reached zero.

What the researchers found

Using time-to-event methods, the authors estimated that 36.2% of participants met the primary outcome by 24 months and 50.1% by 36 months. For the secondary outcomes, the estimated proportions at 36 months were 56.4% and 26.9%. The summary available does not specify which of the three secondary measures each figure refers to, so readers will need the full paper for that breakdown.

The researchers also looked at people who still had likely long COVID at 18 and 24 months after infection. In these groups, the relative timing of later improvement was similar to that seen in the cohort overall. Having symptoms for longer therefore did not, on this evidence, rule out later improvement.

Interpreting the results

The figures are estimates from a cohort that was followed over time, not results from a randomised trial, so they describe the natural course of the condition and say nothing about the effect of any treatment. Because the cohort was restricted to people who already had likely long COVID at 12 months, the findings do not describe the course of the illness in its first year. A threshold-based definition also means the headline figure depends on where the LCRI cut-off is set. The stricter outcome of a score of zero was reached by a considerably smaller proportion than the primary outcome, so many participants who improved probably still had some symptoms.

Equally, the results imply that roughly half of this group had not crossed the improvement threshold by 36 months. For those people, long-lasting illness remains the reality.

Why it matters for research

Long COVID trials need realistic assumptions about how many participants might improve without any intervention. If about a third of people with persistent illness at 12 months cross the improvement threshold in the following year anyway, an intervention study without a suitable comparison group could wrongly credit a treatment for spontaneous recovery. The authors describe their findings as preliminary data for planning interventional trials aimed at improving or resolving long COVID, and note that a sizable proportion of people show some symptom reduction over time.

The next steps are likely to include identifying which clinical or biological features predict improvement, and testing whether particular treatments speed it up. Trials will also need to choose their outcome definitions with care, given that the share of participants counted as improved differed markedly across the measures used here.

Source: About half have meaningful drop in long COVID symptoms by 36 months, study finds (Medical Xpress, 11 November 2026).

Image: AI-generated illustration.