Evaluation of early antibiotics use in non-severe COVID-19 patients admitted with low risk of bacterial infection

The use of antibiotics is common in the treatment of COVID-19, but adequate evaluation is lacking. This study aimed to evaluate the effect of early antibiotic use in non-severe COVID-19 patients admitted with low risk of bacterial infection. The multi-center retrospective cohort study included 1613 non-severe COVID-19 inpatients admitted with low risk of bacterial infection. During the follow-up of 30 days, the proportion of patients progressed into severe type COVID-19 in the early antibiotics use group was almost 1.5 times than that of the comparision group. In the mixed-effect model, the early use of antibiotics was associated with higher probability of developing severe type, staying in the hospital for over 15 days, and secondary infection. However, it was not signicant association with mortality rate. Analysis with propensity score-matched cohort displayed similar results. It is suggested that antibiotic use should be avoided unless absolutely necessary in non-severe COVID-19 patients, particularly in the early stages. factors caused by may exacerbate hypoproteinemia and secondary brinolysis hyperactivity. Our showed a signicant association between early antibiotics use and drug-induced liver injury. So we cannot rule out the effect of liver function damage on albumin and d-dimer levels. According to the above, the intensication of cytokine release, hypoproteinemia and D-dimer elevation caused by antibiotics may also contribute to the progression of COVID–19. that the rate of secondary bacterial infection during hospitalization in the COVID–19 patients treated with antibiotics early signicantly higher than that in patients who did not (20.42% vs. 12.22%). The possible for this as follow: rst, of gut microbiota, of factors and other mechanisms caused by antibiotic use to human immunity in early admitted COVID–19 patients, risk of nosocomial the of patients’ opportunistic 31 in early COVID–19 may even death in of use of among non-severe COVID–19 strict the of the disease. study provides evidence-based support for optimizing antibiotic use in COVID–19 (DID) of antibiotics on and brinolytic while controlling for confounding in linear regression analysis. DID to examine outcome measures for treatment groups and comparison groups that are not randomly assigned. The DID estimations from linear regression models were to capture the net effects of the early antibiotics use, where a negative or positive estimate from the DID models would indicate that a measure of blood examination indicator decreased or increased more over time in patients EAU group than those in NEAU group. The model for age, comorbidities (hypertension and and treatment (antivirus drugs). All


Introduction
The coronavirus disease 2019  caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has spread rapidly worldwide since December 2019 1 . This pandemic has brought a major challenge to the current global health systems 2 . However, except for treatment experience of earlier strains of coronavirus, severe acute respiratory syndrome (SARS-CoV) and Middle East respiratory syndrome coronavirus (MERS-CoV), there are no speci c treatments against COVID-19 to date 3 .
The use of antibiotics is common in the treatment of COVID-19. A recent review found that bacterial/fungal co-infection was present in only 8% of patients with COVID-19, however, 72% received antibacterial therapy 4 . The possible explanation is that the clinical symptoms of COVID-19 are similar to those of bacterial pneumonia, such as coughing, fever, and fatigue 5 . Moreover, 44.3% of COVID-19 patients showed increase of C-reactive protein 6 . When these disease diagnoses cannot be effectively identi ed, clinicians usually give empirical or prophylactic antibiotic treatments against COVID-19. And some national guidelines and cases series have suggested the use of broad spectrum antibiotics or the bene t of atypical antibiotic cover 7,8 .
Severe COVID-19 is an important cause of death in con rmed patients 9 . However, in fact most COVID-19 patients have mild clinical symptoms in the early stages. A report of 72,314 cases by the Chinese Center for Disease Control and Prevention showed that 81% of COVID-19 patients were classi ed as nonsevere patients 10 . When non-severe patients were admitted, their speci c symptoms with COVID-19 were not obvious, and laboratory con rmation could not be obtained quickly due to the limited ability of nucleic acid testing. Therefore, it is anticipated that during the COVID-19 pandemic an increased number of non-severe patients will require commencement on empirical antibiotic therapy. In viral infections, empirical or prophylactic antibiotic treatment has long been controversial 11,12 . Reliable evidence on whether antibiotic treatment has an impact on progression and outcome in patients with non-severe COVID-19 is required. However, there is a lack of research.
Based on the data of patients admitted with non-severe COVID-19, this study used a retrospective cohort design to analyze the effects of antibiotic use within 48 hours of admission on disease progression, secondary bacterial infections, length of stay, and mortality rate, to provide clinical evidence for the formulation of prescription and management strategies of antibiotic therapy for COVID-19 patients.

Results
Overall, 2501 patients were admitted to 4 hospitals in Hubei Province, China. According to the inclusion/exclusion criteria, after excluding 888 patients, eventually 1613 patients were included in the analysis (Fig. 1). Among them, 996 patients received antibiotics within 48 hours after admission (EAU group) and 617 did not receive antibiotics or received antibiotics exceeding 48 hours after admission (NEAU group). The median age at diagnosis in NEAU group was 57 year-old and 54 in EAU group and the difference was statistically signi cant (P = 0.0028). Around 55% of the patients were female in both groups.
Compared with the NEAU group, the EAU group had higher prevalence of cough and fever, but lower prevalence of hypertension. In Addition, the EAU group had a higher percentage of patients receiving antiviral therapies (96.29% vs. 82.01%; P < 0.0001) than patients in the NEAU group.
Progression to severe type COVID-19 During the follow-up of 30 days, out of the 1613 patients admitted with non-severe type COVID-19, 498 patients progressed into severe type. The proportion of patients progressed into severe type COVID-19 in the EAU group was almost 1.5 times than that of NEAU group (36.24% vs 22.20%; P<0.0001). In the mixed-effect Cox model treating site as a random effect, after adjusting for age, gender, comorbidities and in-hospital medications (antiviral drugs), the early use of antibiotics was associated with higher probability of developing severe type (adjusted HR = 1.87, 95% CI: 1.53-2.29) ( Fig. 2A, Table2).
Further analysis was done with propensity score-matched datasets, in which 1222 patients were included, with 611 patients in NEAU group were matched with 611 patients in the EAU group at a ratio of 1:1. The results remained consistent, showing higher risk of turning into severe type in EAU group (adjusted HR = 1.67, 95% CI: 1.35-2.09) (Fig. 2B, Table2).

Length of stay
The average length of stay in EAU group was 18 days and NEAU group 13 days (P<0.0001). In the mixed-effect model, the use of antibiotics was associated with higher risk of staying in the hospital for over 15 days (adjusted OR = 2.34, 95% CI: 1.88-2.92) (Table2). In propensity score-matched cohort analysis, higher risk was also observed for patients administered with antibiotics within 48 hours after admission (adjusted OR = 2.20, 95% CI: 1.72-2.80).

Secondary bacterial infection
The incidence rates of secondary bacterial infection during 30 days of in-hospital follow-up in the EAU group and NEAU group were 20.42% and 12.22% respectively (P = 0.0007). The mixed-effect model showed that the patients in the EAU group was at higher risk of experiencing secondary infection (adjusted OR = 1.90, 95% CI: 1.32-2.75) (Table2). Analysis with propensity score-matched cohort displayed similar results (adjusted OR = 1.69, 95% CI: 1.14-2.51). Liver function, kidney function and brinolytic activity As shown in Table 3, adjusting for patients' sex, age, comorbidities, and whether or not received antivirus drugs), the DID estimator was negative and statistically signi cant at the 5% level in ALB and A/G, and it was positive and statistically signi cant in ALT, UA, and D-Di. The albumin and A/G decreased by 1.5 g/L and 0.1, respectively; while ALT, UA, and D-Di increased by 22.6 U/L, 23.6 μmmol/L, and 1.9μg/mL FEU, respectively. The effect of early antibiotics use on other blood examination indicators was not signi cant in this study.

Discussion
Our study found that in the absence of clear evidence of bacterial infection, early empirical or prophylactic antibiotic treatment augmented the risk of progression from non-severe to severe, increased the incidence of secondary bacterial infections, and prolonged hospitalization in non-severe COVID-19 patients. In addition, there was no signi cant difference in mortality rate between the non-severe COVID-19 patients who received antibiotics early and those who did not.
The composition of balanced gut microbiota has a signi cant in uence on the effectiveness of lung immunity 15 .Disruption of gut microbiota has been shown to impair pathogen clearance capability in the lung and increase susceptibility to in uenza virus infection in lungs 16 . Coincidentally, the SARS-CoV-2 receptor ACE2 is highly expressed on human small intestinal enterocytes, but small intestinal enterocytes can be readily infected by SARS-CoV and SARS-CoV-2 17,18 .
Previous studies showed that antibiotic use in SARS-CoV-2-infected individuals may exacerbate dysbiosis of the gut microbiota and contribute to further lung injury 16,19 . Thus, the imbalance of gut microbiota caused by early antibiotic use might have a potential impact on the progression of non-severe COVID-19 patients.
Studies warned antibiotic treatment without clear evidence of bacterial infection might lead to cytokine storms and septic shock 20 . Antibiotics themselves can stimulate the immune cells to secret pro-in ammatory cytokines (such as TNF-α, IL-1, IL-2 and IL-6), and increase the Toll-like receptor 4 (TLR4) expression [21][22][23] . It means that the release of pro-in ammatory factors due to antibiotics may contribute to cytokine storm which accompanies COVID-19. The systemic in ammatory response syndrome (SIRS) caused by the release of a high number of pro-in ammatory cytokines leads to vascular system damage and extensive microthrombosis [24][25][26] . Moreover, some pro-in ammatory factors such as IL-6 and TNF-α inhibit the synthesis of albumin in the liver [27][28][29] .
Albumin is involved in the regulation of coagulation function, and hypoproteinemia can lead to hypercoagulability 30 . In our study, the use of antibiotics resulted in a decrease in albumin and an increase in D-dimer, which indicated that increased release of pro-in ammatory factors caused by antibiotics may exacerbate hypoproteinemia and secondary brinolysis hyperactivity. Our study also showed a signi cant association between early antibiotics use and druginduced liver injury. So we cannot rule out the effect of liver function damage on albumin and d-dimer levels. According to the above, the intensi cation of cytokine release, hypoproteinemia and D-dimer elevation caused by antibiotics may also contribute to the progression of COVID-19.
This study also found that that the rate of secondary bacterial infection during hospitalization in the COVID-19 patients treated with antibiotics early is signi cantly higher than that in patients who did not (20.42% vs. 12.22%). The possible reasons for this as follow: rst, imbalance of gut microbiota, release of pro-in ammatory factors and other mechanisms caused by antibiotic use damaged to human immunity in early admitted COVID-19 patients, which could increase the risk of nosocomial bacterial infection; second, the invasion of patients' own opportunistic bacteria was also one of the reasons for secondary bacterial infection. A study reported that almost all of COVID-19 patients who died were complicated by secondary bacterial infection 31 .Thus, reducing empiric or prophylactic antibiotic treatment in early admitted COVID-19 patients may reduce the risk of secondary bacterial infection even death in these patients.
Our study provided a timely evaluation of the use of antibiotics among non-severe COVID-19 patients with a strict inclusion and exclusion standard. It was the rst to put a focus on the transformation of the severity of the disease. This study provides evidence-based support for optimizing antibiotic use guidelines in COVID-19 patients. The study also has some limitations. First, as this study was a retrospective study, the missing of some parameters might lead to the deviation of our observation results. For example, the absence of white blood cell counts and procalcitonin (PCT) data in 404 patients after 48 hours may lead to a bias in secondary infection rates. Second, in this study, we were unable to retrieve pre-hospital self-medication, especially the use of antibiotics, from the electronic medical record system, so we could not observe their effect on the changes in patients' condition and prognosis. Third, according to our data, 62 antibiotics, belonging to 16 major classes of antibiotics, were used in the currently enrolled COVID-19 patients from four hospitals in Hubei province. Due to diversity of the types of antibiotics used clinically and the different courses of treatment, we were unable to speci cally assess the potential impact of a particular antibiotic. Forth, there are regional and ethnic limitations. Large clinical cohort studies targeting different regions and ethnic groups are needed to further explore the impact of early antibiotic use on COVID-19 patients.

Conclusion
This study found that early empirical or prophylactic antibiotic treatment against non-severe COVID-19 patients is signi cantly associated with the risk of progression from non-severe to severe, secondary bacterial infections, and prolonged hospitalization. Furthermore, non-severe COVID-19 patients received antibiotics was more prone to Hypoproteinemia and D-dimer elevation. Regarding the above-metioned effects of antibiotic use, we suggest that antibiotic use should be avoided unless absolutely necessary in non-severe COVID-19 patients, particularly in the early stages.

Ethical Statement
This study was approved by the Medical Ethics Committee of Tongji Medical College, Huazhong University of Science and Technology. The requirement for informed consent was waived by the Ethics Committee. Only pseudonymized data with no risk of identi cation were used for our analyses. In this study, early antibiotic use was de ned as patients receiving antibiotic treatment within 48 hours after admission. Since COVID -19 patients and bacterial pneumonia patients all have clinical symptoms such as cough, fever, elevated C-reactive protein and pulmonary imaging changes, patients with white blood cell (WBC) count<9.5*10^9/L (3.5-9.5*10^9/L) and procalcitonin (PCT) < 0.5ng/ml (0-0.5ng/ml: a low rate of bacterial infection) within 48 hours after admission were de ned as having a low risk of bacterial infection. Therefore, the inclusion criteria contained: 1) patients with SARS-CoV-2 infection who were admitted to the above-mentioned hospitals in Hubei, China from 31 st December 2019 to 31 st March, 2020; 2) within 48 hours after admission, respiratory rate (RR) <30 breaths/min, pulse oxygen saturation (SpO2) > 93%, without shock or acute organ failure. The exclusion criteria contained: 1) patients who were intubated, dead, or discharged within 24 hours of admission; 2) within 48 hours after admission, white blood cell (WBC) count 9.5*10^9/L or procalcitonin (PCT) 0.5ng/ml; 3) received antibiotics treatment within 48 hours after admission, but the treatment course was less than 3 days.

Study design and patients
The demographic information, clinical symptoms, medical history, in-hospital medication, and clinical outcomes were obtained from the electronic medical system. Laboratory data (WBC count, PCT, aspartate aminotransferase (AST), alanine aminotransferase (ALT), albumin (Alb), albumin /globulin ration (A/G), serum creatinine (Scre), blood urea (BUN), uric acid (UA), D dimer (D-Di)) were collected from the laboratory information system. The personal identi cation information including name and ID was anonymized and a new study ID was generated for each patient to avoid the possibility of identifying individual patient.

Exposure and outcomes
The exposure in this study was de ned as the treatment of antibiotics prescribed within 48 hours after admission, with a course of treatment≥3 days; and patients in this group were classi ed as early antibiotic use group (EAU group). Otherwise, patients were de ned as the non early antibiotic use group (NEAU group). The antibiotics were identi ed using the Anatomic Treatment and Chemical classi cation, code J01. The study outcomes were: 1) progressing from non-severe type COVID-19 into severe type; 2) length of stay over 15 days; 3) secondary bacterial infection (WBC count>9.5*10^9/L or PCT>0.5ng/ml) during 30 days of in-hospital follow-up. 4) all-cause death during 30 days of in-hospital follow-up.

Statistical analyses
Data are presented as the medians and interquartile ranges (IQRs), or numbers and percentages (%), as appropriate. Comparison of parameters between two groups were conducted with the Wilcoxon-Mann-Whitney-Test for continuous variables. For categorical variables, Pearson's χ 2 test or Fisher's exact tests were used. The risk of outcomes of interest was calculated by the Cox proportional hazard model if hazard curves for the EAU and NEAU groups were proportional (determined by the Kaplain-Meier curve) or Logistic regression. Site was modeled as a random effect in the mixed-effect Cox model and random effect logistic regression. Multivariate analyses were all adjusted for age, gender, comorbidities (hypertension and diabetes), and treatment (antivirus drugs). Based on potential confounding factors associated with the exposure to antibiotics, including age, gender, comorbidities (hypertension and diabetes), and symptoms (cough and fever), propensity score-matched cohorts were established. The EAU group and its matched control unit was set to have the same proportion of female/male patients.
The difference-in-difference (DID) methodology was employed to evaluate the impact of antibiotics use on liver function, kidney function, and brinolytic activity, while controlling for confounding factors in linear regression analysis. The DID approach has been widely used to examine outcome measures for treatment groups and comparison groups that are not randomly assigned. The DID estimations from linear regression models were able to capture the net effects of the early antibiotics use, where a negative or positive estimate from the DID models would indicate that a measure of blood examination indicator decreased or increased more over time in patients EAU group than those in NEAU group. The model were also adjusted for age, gender, comorbidities (hypertension and diabetes), and treatment (antivirus drugs). All analyses were performed using SAS 9.4 (by SAS Institute Inc., Cary, NC, USA).

Declarations Data availability
The data that support the ndings of this study are available from the corresponding author upon request.    Kaplain-Meier Curves for cumulative probability of progression to severe type COVID-19 during 30-day follow-up duration in early antibiotics use/non-early antibiotics use cohort among 1613 patients in unmatched and 1222 matched cohort