We performed a detailed assessment of 1,721 persons with history of SARS-CoV-2 five to 17 months after the verified acute infection and compared to 14,388 controls. A wide variety of symptoms associated with prior infection and for objective measures associations were observed for smell, taste, grip strength, and memory recall.
Our case population represents the severity spectrum of the acute SARS-CoV-2 infection, ranging from no symptoms to severe illness with 5% hospitalized in the acute phase.2,26,28 It follows that our study population consists mainly of persons who did not require hospitalization during the acute illness. The comorbidities that predisposed to severe acute infection in our sample were the same as others have reported17, suggesting that our sample is representative of those infected with SARS-CoV-2. Based on self-report, one fifth had not recovered and 1% still suffered severe symptoms 13 months after the infection.
Deficits in smell and taste are common symptoms of the acute SARS-CoV-2 infection29 and reports suggest full recovery in most at six months.30,31 We found both subjective and objective measures of smell and taste impairment to be more common among cases than controls in our study, with slow temporal improvement of symptoms. Some tests of smell and taste improved with time, with hyposmia normalizing at nine to 10 months after infection, but partial anosmia and partial ageusia did not improve.
Sensorineural hearing loss is a recognised complication of viral infections and there are multiple reports on hearing loss in persons with history of SARS-CoV-2, with most studies based on self-reported questionnaires or medical reports without conclusive hearing tests.32 Here, cases noted worsening of hearing from before the pandemic four times more often than controls, with half of them linking the noted change to the infection. However, objective hearing measures did not associate with history of SARS-CoV-2. Thus, we do not have objective evidence of SARS-CoV-2 causing hearing loss among our mostly non-hospitalized patients.
We observed lower grip strength in persons with prior SARS-CoV-2 infection, likely due to deconditioning. Grip strength, a measure of muscle strength, is a strong predictor of cardiovascular disease and mortality,33 and indeed incident cardiovascular outcomes appear to be more common in survivors of acute Covid-19 than controls.34 It should be noted that although significant, the difference in grip strength between cases and controls was small. We did not observe association between prior infection and exercise capacity as measured by a cardiopulmonary exercise test.
It has been suggested that chronic myocardial inflammation is a common complication of SARS-CoV-2 infection, irrespective of both pre-existing conditions and severity of the acute infection.35–37 One case-control study, assessing 443 individuals after SARS-CoV-2 infection, reported a small reduction in left ventricular ejection fraction and higher concentration of hs-TNT and NTproBNP after infection compared to controls.37 Complicating the interpretation of that study is the lack of contemporary controls, as the control data were derived from persons assessed prior to the pandemic, precluding exploration of time trends in measures as well as any effect of the pandemic itself. The lack of association of the cardiac biomarkers hs-TNT and NT-proBNP with history of SARS-CoV-2 infection in our study of 1721 cases, argues against a persistent myocardial involvement. Similarly, we found no evidence of persistent systemic inflammation, hematologic abnormalities, kidney or liver dysfunction using conventional blood biomarkers.
We performed extensive cognitive testing and observed that poorer delayed and immediate memory recall associated with history of SARS-CoV-2 infection, but the associating effects were small. We calculated the prevalence of impairment in delayed memory recall as 13.4% for cases and 7.5% for controls providing an objective measure of the memory disturbance commonly reported after infection. The prevalence of memory recall impairment in our study is comparable to the 12% reported by Becker et al25 for non-hospitalized persons. However, we did not find significant differences between cases and controls in other cognitive tests, unlike Becker and colleagues who reported high prevalence of cognitive impairment of many domains in the smaller study of 740 persons evaluated after Covid-19 in a clinical setting.
The protracted symptoms of fatigue and neurocognitive disturbance after SARS-CoV-2 infection are reminiscent of other post-infective fatigue syndromes and myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS).38–40 The subjective cognitive impairments that are some of the more debilitating symptoms of ME/CFS41 have been captured by objective measures42 but discrepancies between symptoms and test results are common, possibly due to inability of cognitive tests to capture mild impairment and deficits affecting real-life tasks.43 The WMS Logical Memory test may be a relatively sensitive measure to detect cognitive deficits following SARS-CoV-2 infection, as it has previously been used to detect subtle changes in memory.44 Furthermore, the pathogenesis of these symptoms are unclear. A study leveraging longitudinal brain imaging data from the UK Biobank reported changes in brain structure associated with prior infection with SARS-CoV-2 but changes in areas related to memory did not associate with cognitive tests results.45 It is also notable that while symptoms of neurocognitive disturbance associate with severity of the acute infection in our study, measured deficits in memory recall do not.
Higher BMI, higher triglycerides and lower HDL-C levels associated with severity of the acute infection but not with prior SARS-CoV-2 infection per se. These are all risk factors of more severe acute infection,46 and thus the observed association can be assumed to reflect the predisposition, rather than being a consequence of more severe infection. The causal relationship between the association of increasing levels of anxiety and depression with more severe infection, seen by us and previously described by others,37 in light of the lack of association of these phenotypes with prior infection, is less clear.
The high prevalence of symptoms among cases compared to controls contrasts notably with the small difference observed in test measures between the two groups, as well as with discrepancies between some symptoms and related test measures. For example, tachycardia was a prominent symptom after infection but there was no significant difference in measured heart rate between the two groups. Cases more often than controls reported having gained weight since before the pandemic, but there was no difference in BMI between cases and controls or in longitudinal measures for cases. Similar observations for hearing are described above. Measured memory impairment was 1.91-fold more common in cases than controls while self-reported memory disturbance was described 3.5-fold more commonly by the cases. These observations support an element of response bias47 in self-reported symptoms following SARS-CoV-2 infection and a more complicated biological or biopsychosocial contribution to the persistent symptoms48 that are not well captured by conventional tests. These are important considerations for both research and clinical assessment of post-Covid conditions. Conventional clinical assessment would thus not be expected to be particularly informative in relating reported symptoms to a past SARS-CoV-2 infection.
Our attempt to estimate the prevalence of long Covid highlights not only how common the symptoms of long Covid are in the general population but also the importance of control data for comparison. The excess of cases meeting our criteria for long Covid was 15% with half of those reporting impact on everyday function, translating to a long Covid prevalence of 7–8%. These estimates do not account for potential biases in self-reported symptoms.
This study has limitations. First, although half of Icelanders diagnosed with SARS-CoV-2 infection before February 2021 participated in the study, participation bias cannot be excluded and it is plausible that cases with more pronounced symptoms were more likely to participate, although demographics and comorbidities were similar among those who participated and those who did not. Second, while the study represents adults of all ages, it does not include children. Third, while the availability of measures before and after the infection with similar longitudinal measures for controls is a particular strength of the study, this sample set was relatively small.
We believe that the inclusion of both historic and contemporary controls is a major strength of our study, allowing for consideration of possible time effects, i.e., general consequences of the pandemic itself (social isolation, reduced mobility) in addition to direct effects of the SARS-CoV-2 infection (viral invasion, resulting illness).
In conclusion, in our comprehensive case-control study of mostly non-hospitalized Icelanders, multiple and diverse symptoms were more common among cases than controls five to 17 months after SARS-CoV-2 infection while objective differences between cases and controls were few. Cases performed worse than controls in tests of smell and taste with improvement in some of these measures over time. Cases also performed worse in tests of grip strength and immediate and delayed memory recall, but differences between cases and controls were small. We show that many symptoms associated with prior SARS-CoV-2 infection are common in the general population and, accounting for that, estimate the prevalence of long Covid to be 7–8%. Discrepancies between symptoms and objective measures suggest an element of response bias in self-reported symptoms and a more complicated biological or biopsychosocial contribution to symptoms related to prior infection than is captured by conventional tests.