From January 20 to April 10, 2020, all COVID-19 inpatient with at least three SARS-CoV-2 laboratory results were eligible and recruited for this study from Renmin Hospital of Wuhan University, which is one of the major tertiary teaching hospitals and government-authorized hospitals for COVID-19 patients.
Diagnosis of COVID-19
All COVID-19 patients were diagnosed and clinically graded according to the interim guidance from WHO  and National Health Commission of China . The diagnosis of COVID-19 was made based on a positive result of nasopharyngeal swab by respiratory pathogen nucleotide test with RT-PCR . According to clinical symptoms, imaging findings and laboratory results, COVID-19 patients were classified into four types: mild, moderate, severe and critical .
Clinical data of patients were obtained from the hospital electronic medical record system, including demographics, severity of COVID-19, comorbidities (including respiratory diseases, cardiovascular diseases, cerebrovascular diseases, kidney and urinary diseases, digestive system diseases, liver and gallbladder diseases, metabolic diseases, immune diseases, nerve and mental system diseases, tumors and operation history), laboratory examinations, anti-virus treatments, prognosis and inflammatory and immunologic markers.
Definitions of returned-positive patients
Returned-positive patients were defined as patients with two consecutively negative RT-PCR results of SARS-CoV-2 tests followed by a positive result (a certain sequence of results: negative, negative, positive), with an interval of at least 24 hours between RT-PCR tests. The rest of the enrolled COVID-19 patients were included in the control group.
The COVID-19 laboratory test assays were conducted according to the WHO recommendation. RNA was extracted and tested by RT-PCR with COVID-19 specific primers and probes. The virus genome sequence was examined in samples of bronchoalveolar-lavage fluid by one of three methods: Sanger sequencing, Illumina sequencing, or nanopore sequencing.
Specific antibodies against IgG and IgM were detected using Human SARS-CoV-2 IgG and IgM Chemiluminescence Analysis (CLIA) Assays panel (Shenzhen YHLO Biotech Co., Ltd., Shenzhen, China) and the high-speed CLIA system iFlash 3000 (Shenzhen YHLO Biotech Co., Ltd., Shenzhen, China). Proinflammatory cytokines including interleukin (IL)-2, IL-4, IL-6 and IL-10 were detected using Human Cytokine Standard Assays panel (ET Healthcare, Inc., Shanghai, China) and the Bio-Plex 200 system (Bio-Rad, Hercules, CA, USA) according to the instructions. NLR was calculated by dividing the absolute neutrophil count by the lymphocyte count.
Descriptive analyses were used to summarize the patients’ epidemiological and clinical features. Continuous variables were presented as the median and interquartile range (IQR) as appropriate, and categorical variables were expressed as absolute numbers and percentages. Chi-squared tests or Fisher’s exact tests were adopted for group comparison of categorical variables.
From the date of confirmed diagnosis of SARS-CoV-2 infection, the course of disease (days) was ordinally categorized into phases by an interval of 5 days. Linear mixed effect models were used to analyze the difference in trends of laboratory markers over time (by phase) between returned-positive patients and their counterparts. We entered treatment, gender, age, disease conditions and antiviral therapy (with interaction term of group and phase) as fixed effect into the model. Patients ID was intercepted as random effects. The significance of group differences in the overall trends of the biomarkers was reported and margins effects (means and 95% confidence interval) were plotted over phases accordingly. All analyses were performed using STATA 16.0 software (Stata Corporation, College Station, TX, USA). A two-sided p value less than 0.05 was considered statistically significant.