Follow Up of COVID-19 Features in Recovered Adults without Comorbidities– Persistent Symptoms and Lab-Abnormalities

Julian Varghese (  medizininformatik@uni-muenster.de ) University of Münster https://orcid.org/0000-0002-7206-3719 Sarah Sandmann University of Münster https://orcid.org/0000-0002-5011-0641 Richard Vollenberg University Hospital Münster Kevin Ochs University Hospital Münster Inga Schrempf University Hospital Münster Christopher Froemmel University Hospital Münster Martin Dugas University of Münster https://orcid.org/0000-0001-9740-0788 Hartmut Schmidt University Hospital Münster Phil Tepasse University Hospital Münster


Introduction
COVID-19 (Coronavirus disease 2019) disease has spread as pandemic, causing more than 1 million global deaths as of Novemer 2020 1 and shows a multifaceted condition with various symptoms due to various organ manifestations [2][3][4][5] . There is also an increasing number of recovered patient cases, which can provide new essential insights on the disease course. While the clinical feature characterization of COVID-19 is well-studied for the disease onset 3,6−8 , follow-up information on persisting symptoms and lab ndings is scarce 9 . In addition, most work on symptomatology has not studied clusters of cooccurring symptoms in a single patient. Indeed, a systematic review by Struyf et al 10 found that combinations of symptoms were not assessed in a single study. However, as newer publications are quickly emerging, we found one study on frequent symptoms clusters 11 . This work is a retrospective evaluation of university hospital data of patients, who recovered from COVID- 19 and were screened for convalescent plasma donation eligibility. Since patients with comorbidities were excluded, our study provides insights on the healthy population, in which COVID-19 disease course is not well-studied.
Standardized symptom characterization report forms and all free-text elements of discharge letters were added for analyses. This facilitates the extraction of infection-related symptoms, which are not explicitly listed in structured forms. Moreover, details on persisting symptoms and duration to symptom-free phase could be su ciently extracted in almost all of the cases. As a simple, but highly expressive visualization technique, network analysis illustrates clusters of symptoms and persisting symptoms in one graphbased plot. Blood sampling enabled comprehensive lab-data including whole blood analysis, serum chemistry and immunoglobulins. The combination of detailed symptom duration documentation and raw lab-data enables deep analyses of lab differences -also within the physiological range -in patients with versus without persisting symptoms. By doing this, the study evaluated lab abnormalities including but not limited to general infection parameters, in particular lymphopenia, being discussed as a hot-topic and prognostic factor in disease onset [12][13][14] .

Setting
This is a single center retrospective study in the University Hospital of Münster in Germany. Patients who recovered after con rmed diagnosis of COVID-19 were invited upon an open public call in the Münsterland region of Germany. The primary purpose of this call was to nd eligible patients for a plasma donation study (local ethics committee approval: AZ 2020-210-f-S) 15 . All included patients provided informed consent for scienti c analyses. Figure 1 illustrates the screening process and data collection. Patients with comorbidities or active infection symptoms within the last four weeks were excluded. Final assessment was carried out via outpatient visit at the University Hospital with medical history taking and lab testing. From this stage, further patients were excluded from this study if comorbidities were deemed to be existing or if there was missing information on symptom onset or symptom duration.

Data extraction
Symptoms were obtained from structured clinical forms and discharge letters. Attending physicians were required to document symptoms onset, symptoms duration and if symptoms are currently existing. Symptoms were de ned to be persistent if they endured for at least 28 days from symptom onset. Single symptoms and possible combinations were extracted, but not the duration of each single symptom as this was not documented consistently.
Lab values and corresponding reference ranges were queried from the laboratory information system. Discharge letters were examined by a physician with infection-related symptoms being mapped to medical concepts of the standardized forms or added as new symptoms. Symptoms being highly similar were summarized to one medical concept.

Data Analysis
Network analysis of symptoms was conducted using the R-package igraph 1.2 16 to show frequency and graph-based co-occurrence of onset-symptoms and persisting symptoms. Descriptive statistics were calculated for patient cohort characteristics. Age and gender were compared in the two cohorts with vs without persistent symptoms using 2-tailed t-tests (alpha=0.05) and Fisher's exact test, respectively. Lab data was summarized with descriptive statistics and proportions below and above reference ranges. In addition, lab values were compared between the two cohorts using 2-tailed t-tests (alpha=0.05; adjustment for multiple testing using Bonferroni correction). Calculations were conducted in R, version 4.03.

Patient Cohort
After participation in the online-survey, 122 patients were screened eligible and attended the outpatient clinic for nal assessment during June and September 2020. Six patients were then excluded due to insu cient information on symptom duration or comorbidities that were found to be existent (n = 116). The age range was 18-69 years, with median of 41 years (IQR 30-54). Table 1 summarizes further cohort details. Comparing the two patient cohorts with (n = 92) vs without persistent symptoms (n = 24), no signi cant difference can be observed regarding age (t-test, p = 0.2497) and gender (Fisher's exact test, p = 0.3415).Follow-up ranged from 22 to 102 days. Figure 2 illustrates follow-up times from symptom onset and symptom durations of all patients.

Symptom Analyses
After extraction of infection-related concepts in the discharge letters, the concept "Reduced Physical Resilience" was subsumed to "Fatigue". Myalgia and signs of arthralgia were subsumed to Myalgia. Tinnitus was added as new symptom as it did not occur in the structured standard. Figure 3 shows the network analysis of onset and persisting symptoms. The ve most frequent onset symptoms were Cough (50%), Anosmia (47%), Fatigue (45%), Fever (42%), Myalgia (28%) and Headache (28%). Among patients with persisting symptoms (n = 24), the three most frequent persisting symptoms were Fatigue (n = 13, 54%, min = 28d, max ≥ 94d), Dyspnea (29%, min = 25d, max ≥ 102d) and Anosmia (25%, min = 41d, max ≥ 91d). A web-based platform was established to upload anonymous symptoms data to generate and reproduce network analysis based on own or other cohorts 17 .
Lab Abnormalities  After adjusting for multiple testing for all lab measurements, IgA was the only parameter -also among other Immunoglobulins -that remained signi cantly different between patients with and without persistent symptoms (see Table 2). It was signi cantly reduced in patients with persisting symptoms (p = 0.0219).

Discussion
This study found that a noticeable amount of COVID-19-recovered patients still had (n = 24, 21%) persisting symptoms. In 13% of the cases they persisted for 60 days and longer. The difference should be explained by our comorbidity-free and relatively young population (median age: 41 ys). Moreover, the majority of the population in Calvao-Schneider et al. had an initial hospitalization rate of 35% vs. 9% in our study.
The main strength of our study is the integration of detailed symptoms duration and raw lab-data, which enabled not only assessment of out of range values but also differences within the physiological range.
The study results show that some of the infection-related lab values remained out of range during followup. In particular, lymphopenia -being frequently discussed as a result of direct or indirect viral interactions with lymphocytes 12-14 -remained in a considerable part of our study population, even beyond 90 days of follow-up (see Fig. 4). To the best of our knowledge, this is the rst COVID-19-related study on the duration of lymphopenia along with other routine lab values going beyond 30 days. Further follow-up studies should assess lymphopenia with additional lab analysis to understand the pathogenetic details of lymphocyte-virus interaction.
The comparison of patients with vs without persisting symptoms did not show any differences of lab values except for one: serum concentration of total IgA antibodies. In comparison, there were no signi cant changes regarding total IgG, IgM and all other lab-tests in Table 2. Secretory IgA is an important factor of mucosal immunity for neutralization of toxins and pathogenic microbes 20,21 . While a high level of serum IgA and SARS-CoV-2-speci c IgA on diagnosis can initially correlate with disease severity 22,23 , our ndings suggest that high levels during or after recovery is associated with a decrease in persistent symptoms. In this regard, IgA may play a more important role than other immunoglobulin types. As this study only reports on lab-value correlations, further research can complement our ndings regarding causality.
The principal limitation of this study is the retrospective single-center design with symptom assessment not being objective and being prone to recall bias. Moreover, the different follow-up times also lead to different time distances between last day of symptoms and the actual lab-assessment. As some labvalues return to normal levels over a certain time, some lab tests and association with symptom persistence could have faded out and thus failed to show signi cance in our analysis.

Declarations
Author Contributions JV established the study design, preprocessed the data and wrote the manuscript. SS conducted data analyses. RV, KO, IMS, CF and PRT were involved in data acquisition and preparation. MD and HHS supervised the project. All authors signi cantly contributed to the manuscript and approved for publication.