2.1 Patient Cohort

Between May 2021 and February 2023, 175 consecutive adult patients (≥ 12 weeks after diagnosis of SARS-CoV-2 infection) were recruited (100% of patients visiting the clinic) in a monocentric study in the PCC outcome clinic for neurological PCC symptoms at the Department of Neurology of the Johannes Kepler University, Kepler University Hospital Linz, Austria. Patients came to the clinic upon referral of a general practitioner or other specialist and had generally undergone various clinical assessments and diagnostic tests prior to the visit in our clinic. The referral criteria were consistent with the WHO definition of PCC, requiring an acute SARS-CoV-2 infection that occurred at least 12 weeks prior and ongoing symptoms without apparent alternative explanation (World Health Organization 2022). Furthermore, neuropsychiatric symptoms had to be the dominant complaint. No further restrictions applied. All patients had confirmed SARS-CoV-2 infection. Patients were not included if they had pre-existing myalgic encephalomyelitis/chronic fatigue syndrome or other severe pre-existing co-morbidities, such as an autoimmune disease requiring constant immune suppression. Sociodemographic data were recorded in all patients. Patients were split into mild/moderate disease course (home-isolated) or severe disease course (hospitalized, ICU). Patients had a median of one (range: 1–6 visits) visit, of which only the first, second, and third visits were included in this study due to high drop-out rates. The first visit was 9.87 (median; IQR = 6.95–13.28) months after the acute infection, and the second one was 3.25 (median; IQR = 1.85–5.01) months later. The last visit was 17.57 (median; IQR = 12.87–22.05) months after the acute infection (11.16 months after the first visit, IQR = 7.93–13.41). Concerning physical therapy, patients were informed about the concept of pacing and, if required, referred to specialized PCC rehabilitation centers. Furthermore, patients were referred to a standardized PCC patient education group led by neuropsychologists of the Department of Neurology, Kepler University Hospital Linz, Austria, and to the outpatient clinic for somatization disorders of the Department of Psychiatry, Kepler University Hospital Linz, Austria.

2.2 Ethics Statement

This study was approved by the Ethics Committee of the Medical Faculty, Johannes-Kepler University, Linz, Austria (1161/2021). This study confirms the World Medical Organization Declaration of Helsinki.

2.3 Clinical Examination

All patients underwent a standardized neurological examination by a neurologist. 127/175 (73%) patients underwent standard laboratory diagnostics, including total blood count, differential blood count, iron metabolism, liver and kidney function parameters, coagulation profile, vitamin B12, folic acid, C-reactive protein, serum electrolytes, thyreotropin, serum glucose, serum cortisol, serum creatinin kinase, and lactate dehydrogenase. Blood sampling mean time was 10:12 a.m. (IQR = 9:00–11:24 a.m.). Further diagnostic procedures (cerebral imaging [CT or MRI], electroneurography, electromyography, EEG, and CSF examination) were performed as needed.

2.4 Symptoms

Clinical neurological examination was performed during Visit 1 and re-performed at every follow-up visit depending on the complaints. Subjective symptoms were recorded during clinical examination. The most common symptoms (≥ 10 patients affected at first visit) were categorized into 13 groups: fatigue, subjective stress intolerance (sStIn), daytime sleepiness, insomnia, cognition deficits, pain (including headache and myalgias), chest oppression, dyspnea, hyposmia, parosmia, tachycardiac episode, vertigo, and tinnitus.

2.5 Questionnaires

Questionnaires regarding sleep/fatigue and sleepiness were administered: The “Fatigue Severity Score” (FSS, cut-off: 36/63 points) (Krupp et al. 1989), “Insomnia Severity Index” (ISI; subthreshold insomnia: 8–14/28 points, moderate clinical insomnia: 15–21/28 points, severe clinical insomnia: 22–28 points) (Morin et al. 2011), and the “Epworth Sleepiness Scale” (ESS, cut-off: 10/24 points) (Johns 1991) were used to determine fatigue, insomnia, and daytime sleepiness and were repeated at the first follow-up. For assessment of memory and cognition, the “Montreal Cognitive Assessment” (MoCA) was performed. The PHQ-D was used to evaluate a broad variety of psychosocial and psychological factors at the first visit. It is well validated in patients with psychiatric and non-psychiatric diseases, in the clinic as well as in research (Fleischer et al. 2022; Gräfe et al. 2004; Spitzer, Kroenke, and Williams 1999). This test evaluates seven different domains of psychiatric disorders: 1. somatization (sensitivity: 77%, specificity: 83%), 2. depression (sensitivity: 75%, specificity: 90%), 3. anxiety (sensitivity: 67%, specificity: 94%), 4. panic disorders (sensitivity: 73%, specificity: 98%), 5. other anxiety/fear states (sensitivity: 89%, specificity: 82%), 6. eating disorders (sensitivity: 58%, specificity: 94%), and 7. alcohol abuse (sensitivity: 57%, specificity: 96%) (Gräfe et al. 2004).

2.6 Impact of Somatization on Subjective Symptoms

According to the ICD-10 catalog, somatization disorders are defined as the repeated presentation of physical symptoms together with persistent requests for medical investigations, in spite of repeated negative findings and reassurances by doctors that the symptoms have no physical basis. If any physical disorders are present, they do not explain the nature and extent of the symptoms or the distress and preoccupation of the patient. This discrepancy as well as the accompanying affective and cognitive components were explicitly elicited during the clinical consultation (ICD 2019). To evaluate the impact of somatization on the symptom load, we calculated the frequency distribution of all recorded symptoms according to somatization status. Furthermore, we evaluated the overall complaint load depending on the somatization status.

2.7 Antibody Titer Measurements

Serum EBV nuclear-1 antigen immunoglobulin G (EBNA-1 IgG) as well as the SARS-CoV-2 IgG against the nucleocapsid protein were measured with a chemiluminescent microparticle immunoassay (CMIA) using the Abbott Architect System EBV EBNA-1 IgG detection kit (Abbott Laboratories, Ref. Nr.: 3P67) and the SARS-CoV-2 detection kit (Abbott Laboratories, Ref. Nr.: 6R86-22) according to the manufacturer's instructions. EBNA-1 IgG is given as the ratio of the relative luminescence intensity of the sample over the control (S/CO). The SARS-CoV-2 IgG is shown as an arbitrary unit (AU) per mL.

2.8 Statistics

Statistics were done with GraphPad Prism (version 6.1). Continuous variables were tested for normality with the D'Agostino and Pearson omnibus normality tests. Friedmann test with multiple comparisons and Dunn's post hoc correction were performed to evaluate the complaint load over time. A Wilcoxon-signed rank test was used to evaluate if score results were significantly higher/lower than the pathological cut-off value (36 for FSS, 15 for ISI, and 10 for ESS). For longitudinal analysis of score results, the Wilcoxon matched-pair signed rank test was used. Differences between hospitalized and non-hospitalized patients were tested with the Mann–Whitney U-test. Fisher's exact test was used to test if symptom frequency differed between somatization positive or negative patients at both timepoints. To evaluate if there is a difference in the number of symptoms (complaint load) between somatization positive and negative, a two-sided t-test was performed. Mediator analyses were done with the DATAtab online statistics calculator (DATAtab e.U. Graz, Austria) (Team Data 2024). To evaluate the effect of various psychiatric diseases on the development of symptoms, we combined all positive results for “depression,” “anxiety,” and “other fear states,” as well as previously diagnosed psychiatric diseases. All non-parametric data are shown as median ± IQR. Parametric data are shown as mean ± SD or min/max for boxplots. Symptom frequencies are indicated as percentages of patients affected and are therefore shown as a single value without error bars. This patient cohort represents a convenience sample. No formal sample size calculation was performed prior to the study initiation.

3.1 Demographic Data

A total of 175 consecutive patients with a median age of 42 years (IQR = 33–52) were included in this study, of which 78% (n = 136) were female. The majority had mild-to-moderate COVID-19 infection, and only 15/175 (9%) were hospitalized or received ICU treatment (4/175, 2%). Median time between COVID-19 infection and the first visit was 9.87 (IQR = 6.95–13.28) months. Although invited, only 92/175 (53%) and 43/175 (25%) had at least one follow-up visit after 3.25 (median; IQR = 1.85–5.01) months and another 11.16 months after the first visit (IQR = 7.93–13.41), respectively (Figure 1).

Overall, 97% of the patients had had SARS-CoV-2 infection confirmed by RT-PCR or lateral-flow antigen tests; the remaining 3% were tested positive for SARS-CoV-2 antibodies. Before the SARS-CoV-2 infection, 108/112 (96%) of the patients were employed or studying at university. Of those, 75 (69%) had not been able to return to work/university in full capacity at first presentation due to residual symptoms (Table 1).

3.2 Clinical Examination and Laboratory Results

At the first visit, 3% of patients reported unspecific neurological symptoms like weakness or hypoesthesia (one patient had intermitting trigeminal hypoesthesia, one patient had hypoesthesia and hypopallesthesia in the lower extremities, one had paraesthesia of the upper extremities, one had dysesthesia of the upper and lower extremities, and one had hypoesthesia of the right side of the face, Table 1). In 48/175 (28%) of patients, a psychiatric disorder was present, of which 28/175 (18%) were newly diagnosed after the infection (Table 1, Figure 2B). In 60/175 (34%) patients, cerebral imaging was performed (e.g., dementia assessment), but none showed pathological findings (Table 1). In 3% of patients, a diagnosis possibly explaining the symptoms ascribed to PCC was made during the assessments: in three patients, small fiber neuropathy was diagnosed (explaining pain and tingly sensations in hands/feet), one patient had an ataxic gait disorder (explaining instability when walking), and one patient had hypothyroidism (explaining fatigue). Serological findings were made in two patients (1%; anticardiolipin antibodies, increased Lupus anticoagulant, and GAD65-antibodies). One patient reporting chest pain had a scintigraphy indicating a corpus sterni contusion. In the routine blood analyses, no specific changes were detected (Table S1).

3.3 Symptoms and Neuropsychiatric Diagnoses at the First Visit

The most common self-reported symptom was fatigue (88%), followed by subjective cognitive deficits (78%), and sStIn (74%). Less commonly reported symptoms included pain, insomnia, and hyposmia (Table 2, Figure 2A).

The symptom load was determined by questionnaires, which were filled out by 132/175 (75%) patients at the first visit. In 125/132 (95%), fatigue was present, and FSS scores were significantly elevated compared to the pathological cut-off (median = 55, IQR = 49–59; p < 0.0001, Figure 3A). Severe clinical insomnia was diagnosed in 29/132 (22.0%) patients, moderate insomnia in 53/132 (40.2%), subthreshold insomnia in 35/132 (26.5%) patients, and 15/132 (11.4%) patients were not affected by insomnia. Median ISI scores were significantly elevated compared to the pathological cut-off (median = 16, IQR = 12–21; p < 0.0001; Figure 3D). Daytime sleepiness was diagnosed in 58/132 (44.0%) patients, and median scores were significantly elevated to the pathological cut-off (median = 11.5, IQR = 6–15; p = 0.04, Figure 3G). When comparing hospitalized patients with mildly affected patients, no significant difference was found in either FSS, ISI, or ESS scores (Figure 3C). Cognitive impairment evaluated by MoCA found impairment in 3/132 (2.3%) of patients, who had no previous diagnosis of dementia, were of average age and had only mild/moderate disease course but suffered from severe subjective cognitive deficits (median = 30, IQR = 30–30, Figure 3F). The PHQ-D questionnaire revealed that 75/123 (61%) patients were positive for somatization, 50/123 (40.7%) for depression, 15/123 (12.2%) for anxiety, 14/123 (11.4%) for panic attacks, 38/123 (30.9%) for other anxiety/fear states, and 1/123 (0.8%) for eating disorder and non-alcohol abuse (Figure 3I).

3.4 Symptom Load at the Second Visit

Only 92/175 (53%) patients had follow-up visits. The most common self-reported symptom was fatigue (70.7%). sStIn still affected 56.5%, and a third (33.0%) reported subjective cognitive deficits (Table 2, Figure 2A).

The questionnaires were filled out in 55/92 (60.0%) of the patients at the second visit. Of those, 37/55 (67.3%) exceeded the cut-off according to the FSS, and median levels were significantly elevated compared to the pathological cut-off (median = 48, IQR = 29.5–58; p = 0.0014, Figure 3A). When comparing the FSS scores of both visits of the 55 patients who filled out the questionnaire twice, a significant reduction in FSS severity was observed (p = 0.0004; Figure 3B).

Severe clinical insomnia affected 3/55 (5.5%) of patients at the second visit, 22/55 (40.0%) had moderate and 18/55 (32.7%) had subthreshold insomnia according to the ISI. Median values were significantly elevated compared to the pathological cut-off (median = 14, IQR = 8–15.5; p < 0.0001; Figure 3D). When comparing the ISI scores of both visits of the 55 patients who filled out the questionnaire twice, a significant reduction in ISI severity can be observed (p = 0.0003; Figure 3E).

The ESS revealed no significant increase of daytime sleepiness at the second visit, although 30/55 (54.6%) had scores > 10/24 (median = 10, IQR 4.5–12.5; p = 0.12, Figure 3G). When comparing the ESS scores of both visits of the 55 patients who filled out the questionnaire twice, no significant difference between the first and second visits can be observed (p = 0.3; Figures 3H).

3.5 Symptoms at the Third Visit

Out of 175, 43 (25.0%) had a second follow-up. Fatigue and sStIn were still the most common symptoms but only present in one-third of the patients. Other symptoms were reported much less frequently or were no longer present (Table 2, Figure 2A). For an overview of the subjective symptom load over time, the median complaint load was calculated. At the first visit, a median of four symptoms were reported (range 1–9). At the second visit, the complaint load had decreased to three (range 0–6), and on the third visit, it significantly dropped to zero (range 0–6, p < 0.0001; multiple comparison correction: first vs. second visit: ns; first vs. third visit p < 0.0001; second vs. third visit p < 0.001; Figure 2C).

Due to the high drop-out rate (only 21/175 [12%] patients filled out the questionnaires three times), no score analyses were performed for the third visit.

3.6 Somatization Increases Symptom Frequency

To evaluate the effect of somatization on the prevalence of PCC symptoms, we split the patient cohort according to the result on the PHQ-D somatization questionnaire (positive or negative). We evaluated the longitudinal progress of symptoms in the two cohorts. sStIn and pain were significantly higher in the somatization-positive cohort at the first visit (Figure 4B,H, Table 3). At the second visit, daytime sleepiness and cognition affected the somatization-positive cohort significantly more (Figure 4D,G, Table 3). The overall complaint load at the first visit was significantly higher in the somatization-positive cohort than in the negative patients (Figure 4L).

As psychiatric diseases and somatization are often interlinked, we evaluated if not only a positive test result on the somatization part of the PHQ-D questionnaire affected the number of symptoms, but if psychiatric diseases act as a modulating factor. To this end, we performed a mediator analysis (Table 4). We could show that although the total effect of somatization on the number of symptoms was statistically significant (c: effect 0.61, p < 0.05), no modulating effect of psychiatric comorbidities was found (c′: effect 0.79, p = 0.1; a and b: effect − 0.18, p = 1.0).

Additionally, we investigated if sociodemographic differences modulate the number of symptoms developed. We repeated the mediator analysis for gender and found that the direct effect (c′) was indeed affected by the gender as the effect increased slightly (effect 0.63, p < 0.05). The indirect effect on the number of symptoms was not significantly altered (Table 4). When using the age as a moderating factor, the direct effect (c′) remains significant, however with a slightly lower effect (effect 0.59, p < 0.05). The indirect effect was not significantly altered (Table 4).

3.7 No Aberrant Immune Activation

Median cortisol level was 12 µg/dL (IQR = 9.0–15.3 µg/dL), which is within the physiological range of 4.8–19.5 µg/dL (indicated in gray) with 10 outliers (Figure 5A). To evaluate possible viral reactivation of EBV, we analyzed levels of EBNA-1 IgG in the serum at the first and second visits. Although most patients were positive for EBNA-1 IgG, indicating previous EBV contact, antibody titer did not change during the examined time course (Figure 5B). For SARS-CoV-2 IgG, we did find a significant increase of IgG from first to second visit (Figure 5C).