Introduction

Even after 3 months of a healed COVID-19 illness, many patients still experience persistent symptoms with a heterogeneous pattern consisting mainly of respiratory and neuropsychological complaints. Still, otolaryngological, endocrinological and dermatological symptoms are frequently reported1,2,3. The most common symptoms are dyspnoea and fatigue (“head fog”). These are particularly impairing for patients in everyday life. Given the ongoing pandemic, structured studies of this patient collective are essential.

Potential mechanisms leading to long COVID are the persistence of viruses or virus components, autoimmunological processes, metabolic and endocrinological dysregulations, psychosocial factors, microvascular and mitochondrial dysfunction. Long-lasting sequalae have already been described after severe illnesses4,5 or after infections with some distinct pathogens6,7. Different organic changes in the context of long COVID are described with proposed diagnostic workups from different societies8,9,10. The real impact of the sometimes only minor organic changes on the ethiological assignment of the complaints is often limited.

Given the large number of patients recovered from COVID-19, long COVID is a problem that cannot yet be fully grasped. Therefore, performing a senseful risk stratification is essential to identify patients at risk of permanent health impairment and address it diagnostically and therapeutically. A direct association between diagnostics and the variously expressed subjective complaints seems possible, to a limited extent, in daily practice. Therefore, this study aimed to correlate the physical and mental symptom burden of long COVID patients with their actual findings from the somatic evaluation.

Methods

Consecutive patients who presented to the specialized long COVID unit of our university tertiary care centre were included. Patients presented on their own initiative or as a referral from their general practitioner. All patients had a SARS-CoV-2 infection detected by polymerase chain reaction at least three months before their visit. All patients were included independently of their initial disease severity (ranging from asymptomatic to hospitalised courses). As recently published, diagnostic workups followed a strict examination algorithm11. In brief, all patients underwent transthoracic echocardiography, body plethysmography, capillary blood gas analysis (BGA) and a 6-min walk test (6-MWT). This is intended to narrow down or exclude common causes of cardiopulmonary symptoms using widely available basic diagnostic tools. In the case of pathological changes, further diagnostics are indicated, while relevant pathologies can be ruled out in the case of inconspicuous results. This approach, focusing on basic diagnostics, is intended to address the high number of affected individuals and the often expensive and more complex additional diagnostics such as cardiopulmonary exercise testing and cardiopulmonary magnetic resonance imaging.

Also, a survey was carried out using the standardised SF-36 questionnaire to obtain a somatic assessment of the patients (version 1.0)12. This questionnaire consists of eight dimensions intended to depict the physical and mental well-being of subjects: general health, physical functioning, role limitations owing to physical health, role limitations owing to emotional problems, energy/fatigue, emotional well-being, social functioning and pain. The dimensions were summarised in the superordinate variables physical component summary (PCS) and mental component summary (MCS)13. The arithmetic mean from the individual physical measuring scales of the SF-36 questionnaire was calculated (general health, physical functioning, role limitations owing to physical health and pain) to determine PCS as a measurement of subjective physical well-being. MCS as a tool for evaluating subjective mental health was calculated analogously as the mean of the psychological parameters (role limitations owing to emotional problems, energy/fatigue, emotional well-being and social functioning).

The study was approved by the ethics committee of the University of Ulm (approval number 406/20) and conducted in accordance with the principles of the Declaration of Helsinki. Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research.

Statistics

For descriptive analysis, continuous variables were expressed as means ± standard deviations, and categorical values were expressed as numbers and percentages. The study group was dichotomised by the median. Thereby, a group with better PCS (above the median) was compared with patients below the median and with MCS accordingly. The Student’s t-test was used to compare means in continuous variables. Comparisons of categorical variables were performed using the chi-square test. An ANOVA analysis was performed with a post-hoc test (Bonferroni) as appropriate to compare different courses of initial COVID-19. To compare the somatic examination findings grouped according to physical or mental self-perception, a supplementary grouped evaluation according to initial disease severity was performed as an unpaired t-test using bootstrapping (number of samples 1000). A two-tailed p-value of < 0.05 was considered statistically significant for every test except in the case of multiple testing in the post-hoc tests, where an alpha level of < 0.01 was used. Analyses were performed using IBM SPSS Statistics 26 (IBM, Armonk, NY, USA).

Ethics approval

The study was approved by the ethics committee of the University of Ulm (approval No. 406/20).

Consent to participate

Every participant gave written informed consent.

Consent for publication

All authors have actively participated in this work, reviewed the final draft, and consented to its publication. This manuscript has not been published in part or in its entirety and is not under consideration for publication in any other journal. No portion of the text has been copied verbatim from any sources, and all information provided is accompanied by appropriate references. We are aware that it is the authors’ responsibility to obtain permission for any figures or tables reproduced from previous publications prior to the final acceptance of the manuscript and to fully cover any costs involved. The authors have no conflicts of interest to declare.

Results

From February 22 to September 13, 2021, 463 patients were prospectively examined. Three hundred sixty-seven had filled out the questionnaire in full and were available for further analysis (see Fig. 1). The cohort was 47.3 ± 14.8 years old, and 57.5% were women. The most common symptoms at the time of presentation in our long COVID unit were fatigue (51.1%) and dyspnoea (42.5%). The remaining baseline characteristics can be seen in Table 1.

Figure 1
figure 1

Patient enrolment. Shown is the full study population and the patients who had to be excluded due to missing SF-36 questionnaire. The resulting collective was dichotomized by the median of the physical component summary (median 70.6) and the mental component summary (median 69.1), respectively.

Full size image
Table 1 Patient characteristics (n = 367).
Full size table

The values for the eight SF-36 questionnaire categories are shown in Table 2. MCS was slightly but significantly lower than the PCS (64.2 ± 22.4 vs. 67.1 ± 22.6, p < 0.001). The initial illness severity from COVID-19 was associated with the intensity of long COVID symptoms. Each dimension of SF-36 and MCS and PCS were significantly different between hospitalised, symptomatic and asymptomatic/oligosymptomatic courses of the initial COVID-19 disease except for “pain” (Table 2) which was different only in trend. Differences between hospitalised and initially asymptomatic/oligosymptomatic patients remained statistically significant in a post-hoc Bonferroni test. Despite COVID-19 severity, corticosteroid medication was associated with symptom severity in PCS but not in MCS. Patients with corticosteroid intake during COVID-19 had worse PCS than patients without it (8.5% vs 2.6%, p = 0.014). In addition to symptom burden, several examination findings were associated with initial symptom severity. In particular, worse values for total lung capacity, forced vital capacity (FVC), and DLCO were found in patients with initially more severe COVID-19 disease, as can be seen in Table 2.

Table 2 Results of SF-36 evaluation and results of transthoracic echocardiography, 6-min walk test, body plethysmography and capillary blood gas test in dependency of initial disease severity.
Full size table

Patients in the two lower quartiles in PCS tended to be older, female, with a higher body mass index and have hypertension (Table 3). There was an impairment of functional capacity in the meaning of a reduced walking distance in the 6-MWT (495.6 ± 83.7 m vs 549.7 ± 71.6 m, p < 0.001) and higher values on the Borg scale for dyspnoea (3.8 ± 2.0 vs 2.0 ± 1.7, p < 0.001) and exertion (3.5 ± 2.2 vs 1.4 ± 1.3, p < 0.001). In body plethysmography, the forced vital capacity and diffusion capacity for carbon monoxide (DLCO) were significantly worse in the group with impaired physical health in the self-estimation by the SF-36 questionnaire (FVC: 92.1 ± 12.9% of target vs 94.9 ± 11.7% of target, p = 0.030; DLCO: 85.6 ± 14.3% of target vs 94.5 ± 14.4, p < 0.001). There were no differences in the other ventilatory volumes. A significant but in absolute terms only slight difference was also seen in blood gases with lower values for oxygen and carbon dioxide in the group with lower PCS.

Table 3 Cardiac and pulmonary function depending on the assessment of physical health by the SF-36 questionnaire.
Full size table

The differences described were also found grouped by initial disease course, although some results were not statistically significant most likely because of the small subgroups and a high scattering rate. For example, reduced walking distance in the 6-MWT was seen in initially hospitalized patients (485.7 ± 100.1 m vs 525.0 ± 74.4 m, p = 0.342) and outpatients (495.9 ± 79.3 m vs 552.6 ± 72.1 m, p = 0.001) or in asymptomatic/oligosymptomatic patients (514.7 ± 120.2 m vs 542.9 ± 69.6 m, p = 0.453). This was similar in DLCO, where initially hospitalized and asymptomatic/oligosymptomatic patients had numerically worse outcomes in the lower two quartiles of PCS (74.7 ± 17.1% of target vs 88.8 ± 74.7% of target, p = 0.118; and 92.7 ± 19.6% of target vs 96.0 ± 12.0% of target, p = 0.574). Symptomatic outpatients had significantly lower DLCO scores in the two lower quartiles of PCS (86.3 ± 13.2% of target vs 94.5 ± 14.5% of target, p = 0.001).

According to the MCS, similar results were found in the classification (Table 4). There were predominantly women in the more restricted group (69.1% vs 46.6%, p < 0.001). Again, walking distance (502.4 ± 73.3 m vs 543.3 ± 85.2, p < 0.001) and values on the Borg scale were significantly lower in the group with lower MCS. There was a significant difference in DLCO (87.1 ± 16.4% of target vs 93.1 ± 12.9% of target, p < 0.001) and, to a lesser extent, FVC and FEV1. Grouped by initial disease severity, the more restricted group again had worse results. This is reflected in the walking distance in the 6-MWT (hospitalized: 466.7 ± 64.4 m vs 560.6 ± 108.9 m; p = 0.050; symptomatic non-hospitalized: 506.8 ± 70.9 m vs 541.3 ± 86.2 m, p = 0.004; asymptomatic/oligosymptomatic: 505.8 ± 98.5 m vs 548.3 ± 75.7 m, p = 0.125).

Table 4 Cardiac and pulmonary function depending on the assessment of mental health by SF-36. The lower two quartiles (< 69.1 points) were described as impaired, and the upper two quartiles (> 69.1 points) as less impaired physical health.
Full size table

The period since COVID-19 was inversely associated with the severity of symptoms, which was only statistically significant in the PCS classification (p = 0.045). The evaluation of laboratory tests showed significantly higher values for D-dimers in the group with more symptoms of PCS. There was only a trend without significance in the MCS. Values for haemoglobin, glomerular filtration rate, high sensitive troponin t and NT-proBNP showed no relevant differences between groups (Tables 3 and 4).

Evaluation of cardiac function by transthoracic echocardiography revealed no differences between the groups, as defined by the SF-36 assessment. Patients with impaired (physical and mental) health showed the same values for left ventricular ejection fraction (LVEF) and left ventricular global longitudinal strain (LV GLS) as patients with good health in the self-assessment.

Discussion

The main findings of this study are as follows: (1) the severity of the initial COVID-19 disease and female sex are associated with higher symptom burden in the context of long COVID; (2) the extent of physical and mental impairment in the self-assessment using SF-36 as well as the initial disease severity correlate significantly with DLCO and distance in the 6-MWT and (3) there was no correlation of symptom burden to markers of the left ventricular function in transthoracic echocardiography.

A predominance of the female gender has been described in other long COVID cohorts1,14. Immunological and psychological causes were discussed, with a potential overlap. Under the assumption of T-cell abnormalities or autoantibodies leading to long COVID, the reported higher number of T-cells and autoantibodies during and shortly after COVID-19 could be potentially causative for ongoing symptoms15,16. On the other hand, women are more prone to depression and anxiety disorders because of their higher oestrogen levels17,18,19. Since the causes of a long COVID syndrome have not yet been fully understood and a multi-causality is likely, the female predominance cannot be adequately explained.

A positive correlation of pulmonary restriction in follow-up with COVID-19 severity was described earlier, especially in hospitalised patients20,21. Data on mild or even asymptomatic courses are scarce. In a study carried out in Wuhan by Huang et al., 1733 patients were examined six months after the reduced walking distance in the 6-MWT, and diffusion impairment dependent on the in-hospital course of the initial COVID-19 disease has been shown during hospital discharge22. In another trial with hospitalised patients, a DLCO reduction could be associated with female sex and radiological abnormalities in follow-up examinations after 3–12 months21. In contrast, our cohort consisted of all clinical courses, including outpatient and asymptomatic, with a mean follow-up time of six months. We could show a correlation of well-being to functional values of diffusion testing and the 6-MWT. While neuropsychological factors can influence symptoms during the 6-MWT and walking distance, the reduction in DLCO suggests an at least transient organic correlate. Whether the numerical differences in DLCO, some of which are small, are actually causal for the patients' dyspnea remains questionable. Patients with a low DLCO can be rehabilitated well and improvements in DLCO and 6-MWT go hand in hand23. Therefore, our results are of clinical relevance, as they illustrate that rehabilitative therapy of dyspnea and exercise capacity also improves patients' self-assessed physical and mental health; thus, no separate therapeutic approach is necessary for this.

In the longitudinal trial by Steinbeis et al., an improvement in respiratory function has been shown, including DLCO over time20. Considering the possible time course of DLCO and the significantly different observation times of the cohorts after COVID-19 infection in our collective, a possible confounder or coincidence is possible here.

In another trial, a time dependency of symptom burden in the context of long COVID has been shown in 13.3% of participants with symptoms lasting more than 28 days after COVID-19 and only 2.3% for more than 12 weeks1. This is in line with the assumption that most cases of long COVID are another form of post-infectious syndromes, which have also shown time dependency7,24,25. It currently remains unclear to what extent an improvement in lung function can be achieved and by which measures this process can be positively influenced. In a synopsis of the achievable final state and, if necessary, further imaging examinations, a distinction can ultimately be made between functional impairment (“functional” long COVID) and actual organ damage as an expression of a distinct disease (e.g. myocarditis, pulmonary embolism, pulmonary restriction). In this context, the increased levels of D-dimers in comparison between more symptomatic and less symptomatic patients may be of interest. These findings are potential indications of prothrombotic conditions, possibly associated with an increased incidence of pulmonary embolism. However, this condition is only suggestive and should be investigated in further studies.

Long COVID is affecting patients regardless of the initial severity of the disease26. It is pointed to a missing or even inverse relationship between COVID-19 severity and the ongoing symptoms in some studies. No differences in symptom burden and walking distance in the 6-MWT between different clinical courses of COVID-19 have been found in a study by Townsend et al.27. There was a relevant selection bias risk with only 153 out of 487 patients (31%) who accepted the offer for the examination in their trial. In return, there is also the possibility of selection bias in our setting of a long COVID unit. There were apparent differences between the self-assessed symptom severity in our cohort of previously hospitalised, outpatient and asymptomatic/oligosymptomatic patients. Initial disease severity was further associated with changes in body plethysmography (specifically total lung capacity, FVC, and DLCO) in our study. From this, it could be inferred that the association of symptom burden and functional parameters would be rather coincidental in nature. Since patients with higher self-assessed symptom severity also had worse outcomes in the subgroups according to initial disease severity, we assume a combined effect. Because the cohorts of initially hospitalized and asymptomatic/oligosymptomatic patients are very small, our results in these groups can only be considered hypothesizing. Cardiac inflammation signs as part of post-COVID-19 sequelae were frequently reported earlier28,29,30. Most of them showed elevated levels of markers for inflammation in non-invasive tissue characterisation using cardiovascular magnetic resonance (CMR) imaging (T1, T2 and ECV). It was unclear whether these data were attributable to long-lasting symptoms after COVID-19. In our cohort, we found no correlation between more severe symptoms and LVEF or LV GLS. In another trial by Joy et al., CMR scans were carried out in seropositive and seronegative healthcare workers, and no differences in markers for inflammation were found. Since there was only one hospitalised patient because of COVID-19 in their cohort, cardiac involvement or impairment is possibly correlated with the severity of the initial COVID-19 course. Elevated markers of myocardial inflammation with normalisation in a follow-up visit have been shown in another trial with 58 hospitalised patients (36.1% admission to intensive care unit)31. The researchers also carried out functional lung tests, blood samples and cardiopulmonary exercise tests, finding no correlation to the symptoms of the patients. The difference in our data may be a consequence of COVID-19 severity or the fact that symptoms were assessed categorically rather than gradually in our study. Interestingly, they found a significant difference in right ventricular ejection fraction and volume between baseline and a follow-up visit, with an improvement over time, unfortunately not evaluated in our data. This could be a goal for further investigation as the right ventricular strain is easy to assess by transthoracic echocardiography32.

Conclusion

A correlation of long COVID symptoms with the severity of the course of the initial COVID-19 disease and cardiopulmonary function markers is suggested by our data, particularly DLCO and the 6-MWT. A longer-lasting residual due to a more severe COVID course appears likely. Conversely, cardiopulmonary diagnostics do not appear necessary in low-symptomatic initial COVID-19 courses or subjectively less impaired patients. Whether the worsening of DLCO or walking distance in the 6-MWT is causal or coincidental for the symptoms needs to be investigated in further studies. Myocardial injury resulting in a reduction of left ventricular function is not correlated with symptoms of long COVID.

Limitations

Selection bias is likely for our investigation because patients were referred to us on their initiative or from a general practitioner. This leads to the assumption that patients with mild and acute symptoms of long COVID were less likely to come to our unit. A study of larger or further prespecified cohorts would provide the opportunity to make more differentiated statements about changes in diagnostics. In particular, the groups with initially hospitalized and asymptomatic/oligosymptomatic patients were too small to provide reliable results in our study.