Pinpointing the true estimate of long-Covid prevalence from asymptomatic, non-severe, and severe Covid-19.
In one of the most extensive and latest long-Covid meta-analyses to date, scientists Chen et al. at the University of Michigan School of Public Health, U.S. synthesized data from 34 studies and computed that the global prevalence of long-Covid stood at 43% of Covid-positive cases. This number has been cited in various news outlets covering long-Covid.
If long-Covid prevalence is really 43%, that means over 180 million people already have or had long-Covid. (Covid-19 cases surpassed 445 million, with over 6 million deaths, as of early March 2022.) That also means the chance of developing long-Covid is almost 50% if we tested positive for Covid-19. But such long-Covid prevalence is likely overestimated. Let’s see why.
What meta-analyses show
(1) In one early meta-analysis, Lopez-Leon et al. identified 15 studies on long-Covid — defined as having at least one symptom persisting for 14–110 days after SARS-CoV-2 infection — in the literature before 1 January 2021.
Harmonizing the studies’ data showed that 80% of 47,910 Covid-19 cases developed long-Covid. The five most prevalent long-Covid symptoms were fatigue (58% of cases), headache (44%), attention disorder (27%), hair loss (25%), and breathlessness (24%).
But this meta-analysis was performed quite early, only meta-analyzing studies from 2020, so sample sizes are rather small. Much of the sample size of this meta-analysis was driven by one study of 44,779 Covid-19 cases.
Access to Covid-19 tests was also limited during the early pandemic, which may have underestimated Covid-19 cases, thus lowering the denominator and overestimating long-Covid cases.
(2) In another meta-analysis, Iqbal et al. scoured the literature as of 6 March 2021 and meta-analyzed 43 studies, finding that:
- For acute post-Covid (<12 weeks), the pooled prevalence of fatigue, breathlessness, anxiety, concentration impairment, and depression is 37%, 35%, 29%, 24%, and 20%, respectively, of Covid-19 cases.
- For chronic post-Covid (≥12 weeks), the pooled prevalence of fatigue, sleep disturbance, breathlessness, ageusia, anosmia, and chest pain is 48%, 44%, 39%, 18%, 17%, and 17%, respectively, of Covid-19 cases.
(3) In a more recent meta-analysis, Ceban et al. identified and meta-analyzed 74 studies on long-Covid — defined as having the symptom for at least 12 weeks after diagnosis — as of 8 June 2021.
- For fatigue (n=25,268), the pooled prevalence is 32% of Covid-19 cases, which affects adults (32%) significantly more than children (7%), but there were no significant differences between males vs. females, hospitalized vs. non-hospitalized, and <6 vs. ≥6 months of follow-up.
- For cognitive impairment (n=13,232), the pooled prevalence is 22% of Covid-19 cases, which were not significantly different between males vs. females, hospitalized vs. non-hospitalized, and <6 vs. ≥6 months of follow-up. (Differences between adults and children were not computed as only one study accessed cognitive impairment in post-Covid children.)
(4) In the latest meta-analysis, “Global Prevalence of Post-Acute Sequelae of COVID-19 (PASC) or Long COVID: A Meta-Analysis and Systematic Review,” Chen et al. pooled data from 34 studies on long-Covid, as of 12 August 2021, comprising a total of 886,388 Covid-19 cases. Long-Covid is defined as having at least one symptom for at least 28 days after diagnosis.
Pooled long-Covid prevalence is:
- 43% of Covid-19 cases in general: 36% at 30-day, 24% at 60-day, 29% at 90-day, and 51% at 120-day follow-up.
- 57% of previously hospitalized Covid-19 cases
- 31% of non-hospitalized Covid-19 cases
- 49% of female Covid-19 cases
- 37% of male Covid-19 cases
The five most common symptoms were fatigue (23%), breathlessness (13%), memory problems (13%), insomnia (13%), and joint pain (13%).
Chen et al. further calculated that female sex and pre-existing asthma were associated with 1.57-times and 2.15-times increased risks of long-Covid, respectively. They also noted that — among other studies that were not meta-analyzed — other risk factors include more severe Covid-19, a high number of symptoms during Covid-19, old age, and pre-existing medical conditions.
So, based on the four meta-analyses, the prevalence of long-Covid is high — ranging from 30–80% of Covid-19 cases. If we take the latest meta-analysis as the main estimate, it’s 43% of Covid-19 cases.
But these prevalence numbers are missing a key statistic — the background prevalence: how much of it would have occurred regardless of Covid-19.
Background prevalence is also high
One of the main limitations of long-Covid studies is not evaluating new-onset symptoms. As Ceban et al. acknowledged, nearly all long-Covid studies did not clarify whether the symptom in question already existed before Covid-19.
If yes, that means most long-Covid studies are also counting the background prevalence, in addition to the new-onset prevalence of long-Covid.
Even if not, most long-Covid studies also did not have a non-Covid control group, which serves to determine whether the symptom in question would have existed regardless of Covid-19, as well as what is the excess prevalence of long-Covid from the background. I also raised a similar problem in a research review published in December 2021.
As Ceban et al. stated:
“Common methodological limitations were the failure to include a non-exposed group in cohort studies, failure to ascertain whether outcomes were present prior to COVID-19 infection, and a lack of sample size justification in cross-sectional studies.”
“…the incidence of depressive and anxious symptoms in the general population has increased since the pandemic onset (Xiong et al., 2020); fatigue and cognitive impairment may be consequences of chronic stress and/or depression resulting from social and economic challenges of COVID-19, rather than a result of infection, in a proportion of PCS [post-Covid-19 syndrome; long-Covid] patients. It is also noteworthy that social consequences may be exacerbated for infected individuals.”
So, let’s now look at what are the background prevalence and what long-Covid studies with control groups show.
Let’s start with the background prevalence of the most common long-Covid symptom — fatigue. Studies conducted before the pandemic found that:
Next, let’s see the background prevalence of the second most common long-Covid symptom — breathlessness or cognitive impairments.
For breathlessness (medical term: dyspnea):
For cognitive impairment, pinpointing its prevalence is trickier. Studies usually measure mild cognitive impairment (MCI), also known as prodromal Alzheimer’s disease. According to a meta-analysis of 11 population-based studies, MCI has a prevalence of 4.5% among 60–69 years, 5.8% among 70–79 years, and 7.1% among 80–89 years. Another meta-analysis of 41 studies put this number at 12.2% for those over the age of 55 years.
However, long-Covid studies don’t examine MCI specifically but rather symptoms of cognitive issues in general. So it’s hard to know what the background prevalence of cognitive symptoms would be.
For other common long-Covid symptoms such as sleep disturbance, headache, joint pain, anxiety, and depression, their background prevalences are also substantial, hovering at around 10–30%. (But note that clinical disorders like anxiety or depressive disorders are much rarer at <5%.)
So, background prevalence research tells us that, at any given point in time, about 10–30% of the general population has symptoms similar to long-Covid.
The latest estimate on long-Covid prevalence is 43% of Covid-19 cases; i.e., 23% fatigue, 13% breathlessness, 13% cognitive issues, 13% insomnia, and 13% joint pain — all of which have substantial overlap with the background prevalence.
Breaking down the 43% prevalence number to severity status, it’s 31% of non-severe and 57% of severe Covid-19 cases. If we deduct them from the background prevalence of 10-30%, let’s just say 20%, we get the long-Covid prevalence of 11% of non-severe and 37% of severe Covid-19 cases.
What controlled long-Covid research shows
One controlled long-Covid study was the work of Logue et al. (2021) in Seattle, U.S., where 29% and 13.6% of Covid-19 (mostly non-severe) survivors had at least one and three symptoms at 9-month, respectively, compared to only 5% in the control group. But the control group consisted of only 21 healthy people — not really a fair proxy for background prevalence.
This study tells us that using the appropriate controls is crucial. Just because a study has a control group doesn’t mean it’s foolproof. So, let’s focus on studies with proper control groups of sufficiently large sample sizes.
(i) In one of the largest controlled long-Covid studies to date, Caspersen et al. from Norway followed over 70,000 people for over a year. Only 774 people (1%) were diagnosed with Covid-19, and 72,953 people (99%) were not. Both Covid and non-Covid groups were demographically similar in terms of age, sex, smoking status, chronic diseases, BMI, and educational level.
Results revealed that Covid-19 survivors at 11–12-month had:
- 16.3% excess cases of altered smell/taste (16.9% of Covid-19 cases vs. 0.3% of non-Covid cases: 16.9%–0.3% = 16.3% excess).
- 14.6% excess cases of poor memory (18.2% vs. 3.6%).
- 13.6% excess cases of fatigue (17.4% vs. 3.8%).
- 9.9% excess cases of breathlessness (11.2% vs. 1.3%).
- 8.1% excess cases of brain fog (12% vs. 3.9%).
- 7.4% excess cases of reduced lung function (7.7% vs. 0.3%).
- 5.6% excess cases of heart palpitations (7.7% vs. 2.1%).
- 3–5% excess cases of chest pain or headache.
- 2–3% excess cases of cough, anxiety, dizziness, or skin rash.
- No statistically significant excess cases of myocarditis, depression, mood swings, sleep problems, joint pain, muscle pain, fever, or kidney disease.
This study also found that severe Covid-19 was about 2-times more likely to lead to long-Covid than mild Covid. For example, brain fog incidence was 14.2% for severely ill and 5.3% for mildly ill Covid-19 survivors. For breathlessness, the numbers were 18.2% and 5.6%, respectively. In this study, females were also more likely to develop long-Covid, specifically brain fog, fatigue, headache, dizziness, poor memory, and altered smell/taste.
(ii) Another similar longitudinal controlled study from the Public Health England also found that the excess cases of long-Covid symptoms (subtracted from the control group) were about 10–15% at 6–7-month:
- 10–22% excess cases of fatigue, altered taste/smell, short-term memory loss, brain fog, or breathlessness.
- 5–10% excess cases of appetite loss, blurred vision, speech impairments, sleep disturbances, cough, chest pain, or palpitations.
- 3–5% excess cases of trembling, swallowing issues, toe/finger twitching.
(iii) Another controlled study from Sweden reported that 26% of Covid-19 survivors developed at least one long-Covid symptom compared to 9% in non-Covid controls, equating to an excess of 17%, at 2-month. At 8-month, these numbers were 15% and 3%, respectively, an excess of 12%. Altered smell or taste, fatigue, and breathlessness were the most common symptoms.
So, these controlled studies inform that long-Covid is a real problem, with the true prevalence of about 10–20% of Covid-19 cases — not as high as some uncontrolled estimates show.
A few more things to consider
But this overall risk of long-Covid at 10–20% of Covid-19 cases could be lower or higher depending on a few factors: for example, vaccination lowers the risk by about half, whereas female sex and pre-existing diseases heighten the risk.
But considering that (i) we don’t test everyone for Covid-19, (ii) the global estimate of asymptomatic (no symptoms) Covid-19 is 40% of cases, and (iii) people don’t usually get tested unless symptoms manifest, the denominator of positive Covid-19 cases is most likely higher than the current statistics.
As many experts believe, current Covid-19 cases are underestimated, assuming a Covid-19 case means a positive PCR test. Ergo, the true prevalence of long-Covid may even be lower than 10–20% of Covid-19 cases.
Moreover, even when controlled, cohort studies have one major drawback: no randomization of participants. Randomization is the only way to negate the countless potential confounding variables between groups. But it’s unethical to inoculate SARS-CoV-2 or placebo into people at random.
For example, current long-Covid studies cannot rule out the possibility that people who are more likely to develop fatigue may also be more likely to develop Covid-19. If this is true, we may not see much changes from the pre-pandemic background prevalence; for instance, a 20% background prevalence of fatigue may get distributed to 15% in the Covid-19 and 5% in the non-Covid group, given an excess of 10%, but still a 20% background prevalence. No studies have attempted to control for this factor, as far as I know, if it’s even controllable without randomization, however. Plus, we aren’t very sure how much of the long-Covid effects are exacerbated by pandemic-related factors, such as social isolation or financial issues.
Lastly, nearly all long-Covid studies did not study asymptomatic cases. But asymptomatic Covid-19 doesn’t seem to result in long-Covid as often.
A study in California reported that 32% of long-Covid cases — defined as at least one symptom for >60 days — were initially asymptomatic. But 60-day follow-up is rather short, no control group was involved, and this study is still a pre-print released over a year ago.
However, two published studies from Korea and Seattle, U.S., with a longer follow-up time of 6–12 months, reported that 0% of the asymptomatic Covid-19 cases developed long-Covid. Another published study from Italy found that only 5.5% of asymptomatic Covid-19 cases led to long-Covid, compared to 44% of symptomatic cases, at 6-month. But this study has no control group, so that 5.5% figure may just be background prevalence.
Based on all these, I estimate that if you get symptomatic Covid-19, the chances of developing long-Covid — or new-onset fatigue, breathlessness, or cognitive issues— is likely one or two in ten. If your Covid-19 is asymptomatic, it’s almost zero in ten. If your Covid-19 progressed to the severe stage, it’s likely one or two in five.