IL6 inhibition in critically ill COVID-19 patients is associated with increased secondary infections
====================================================================================================

* Lucas M. Kimmig
* David Wu
* Matthew Gold
* Natasha N Pettit
* David Pitrak
* Jeffery Mueller
* Aliya N. Husain
* Ece A. Mutlu
* Gökhan M. Mutlu

## ABSTRACT

**Background** Anti-inflammatory therapies such as IL-6 inhibition have been proposed for COVID-19 in a vacuum of evidence-based treatment. However, abrogating the inflammatory response in infectious diseases may impair a desired host response and predispose to secondary infections.

**Methods** We retrospectively reviewed the medical record of critically ill COVID-19 patients during an 8-week span and compared the prevalence of secondary infection and outcomes in patients who did and did not receive tocilizumab. Additionally, we included representative histopathologic post-mortem findings from several COVID-19 cases that underwent autopsy at our institution.

**Results** 111 patients were identified, of which 48 had received tocilizumab while 63 had not. Receiving tocilizumab was associated with a higher risk of secondary bacterial (50% vs. 28.6%, p=0.021 and fungal (8.3% vs. 0%, p=0.078) infections. Consistent with higher number of infections, patients who received tocilizumab had higher mortality (39.6 vs. 17.4%, p=0.016). Seven cases underwent autopsy. In 3 cases who received tocilizumab, there was evidence of pneumonia on pathology. Of the 4 cases that had not been given tocilizumab, 2 showed evidence of aspiration pneumonia and 2 exhibited diffuse alveolar damage.

**Conclusions** Experimental therapies are currently being applied to COVID-19 outside of clinical trials. Anti-inflammatory therapies such as anti-IL-6 therapy have the potential to impair viral clearance, predispose to secondary infection, and cause harm. We seek to raise physician awareness of these issues and highlight the need to better understand the immune response in COVID-19.

While there has been a dramatic increase in the number of clinical trials, there remains a shortage of effective therapies for COVID-19, particularly for patients who develop critical illness. The management of such patients remains largely supportive. Early reports from China suggested that an exaggerated immune response may play a role in the development of respiratory failure, shock, and multiorgan dysfunction in critically ill patients with COVID-19 1. Similarities between the exaggerated immune response associated with COVID-19 and the cytokine release syndrome (CRS) reported in patients with CAR T-cell therapy led to the use of tocilizumab, an anti-IL-6 therapy to attenuate hyperimmune responses associated with COVID-19 2. However, inhibition of IL-6 may also have adverse consequences. Mice lacking IL-6 response have impaired immunity against viral, bacterial and fungal pathogens 3. Humans treated with tocilizumab had higher risk of serious bacterial, skin and soft tissue infections 4-7. Lastly, IL-6 appears to play a complicated role viral clearance8. To date, several reports detail institutional experiences with treating COVID-19 with tocilizumab.9-13 As expected, inflammatory markers (cytokines, temperature) generally demonstrated a profound response to tocilizumab administration. However, as most reports lack a comparison group, it is difficult to ascertain whether patients clinically benefitted from inhibition of the inflammatory response. Similarly, available reports often restrict secondary infections to documented blood stream infections, which may significantly underestimate the infectious complications of anti-IL-6 therapy in the critically ill. We sought to determine the incidence of secondary infections and outcomes of patients who received tocilizumab for COVID-19 compared to those who did not at our institution.

## METHODS

### Patients

We retrospectively analyzed the COVID-19 ICU admission data between the dates of March 1, 2020 and April 27, 2020 and identified records from 111 patients admitted to the COVID-19 ICU during that period. The study was approved by the University of Chicago Institutional Review Board. Additionally, 7 patients underwent autopsy, the histopathology of which was reviewed with our colleagues from pathology.

### Interventions

Tocilizumab was included in our internal institutional protocol for the treatment of COVID-19, specifically for use in patients with progressive clinical deterioration and elevated inflammatory markers at the discretion of the treating team and infectious diseases consultation service. Our protocol recommended 400 mg flat dosing of tocilizumab with the potential for redosing based on clinical response (e.g. oxygenation status, hemodynamic stability, inflammatory marker response). Tocilizumab was generally not considered in patients with liver function test (LFT) abnormalities, significant cytopenias, or documented ongoing infection. Similarly, patients enrolled in clinical trials (e.g. remdesivir) were not eligible to receive tocilizumab.

### Statistical Analysis

Statistical analysis was conducted with SPSS version 26 (IBM, Armonk, NY) and R version 4 (Package Logistf). APACHE II score and Charlson comorbidity index were not normally distributed and thus were analyzed with Mann-Whitney U-test. Count variables were analyzed with Chi-square and continuous variables were analyzed with t-tests.

Multivariate regression analysis was completed in R with the outcome variable being the development of bacterial infection. The predictors (i.e., independent variables) were selected using Chi-square statistics. Variables that are statistically different between those were given tocilizumab and those who were not, were sex with the predominance of males in the tocilizumab group, history of transplant (n=5, all patients were given tocilizumab), use of immunosuppressive agents (n=8, p=0.074 with a trend towards tocilizumab), and APACHE II scores (see table 1). There was high collinearity between history of transplantation and immunosuppressive use as expected. Thus, we only included immunosuppressive use in the multivariate model. There was a trend towards a higher proportion of patients with diabetes mellitus with organ damage in those with bacterial infection (p=0.054) and past medical history of deep venous thrombosis (DVT) or pulmonary embolism (PE) (n=8, exclusively noted in those who did not have a bacterial infection, p=0.023). No other clinical variables were statistically different by tocilizumab use or the development of bacterial infections. Other clinical variables that are included in Charlson were not statistically significant and therefore were not included in the model. Therefore, the final model included age, sex, tocilizumab, diabetes with organ damage, DVT/PE, Charlson comorbidity index, APACHE II score and immunosuppressive use. Of these, age, sex, diabetes with organ damage, Charlson comorbidity index, and immunosuppressive use were not independently associated with the development of bacterial infections (all p>0.05). The reported odds ratios (ORs) reflect exponent (β) from the Firth regression14. Firth regression was chosen because history of DVT/PE albeit in small number of cases caused separation in the model.

View this table:
[Table 1.](http://medrxiv.org/content/early/2020/07/06/2020.05.15.20103531/T1)

Table 1. Co-morbidities and Outcomes

## RESULTS

We identified 111 patients admitted to the COVID-19 ICU during that period. Of these, 28 received tocilizumab 400 mg once except for three patients who received two doses (800 mg total) and one patient who received a single 800 mg dose. Our protocol recommended 400 mg flat dosing of tocilizumab with the potential for redosing based on clinical response (e.g. oxygenation status, hemodynamic stability, inflammatory marker response). We compared bacterial and fungal infections in those that received the drug to those that did not. There were no differences in patient baseline characteristics (including age, sex, Charlson co-morbidity index (CCI), as well as variables that may not be captured by the latter index such as immunosuppression, hypertension, etc.) between the two groups (Table 1). Secondary infections were defined by positive culture data or high clinical suspicion of infection requiring the initiation of antimicrobials and documentation in the progress note. In the tocilizumab group, only infections occurring after the tocilizumab dose were counted.

Receiving tocilizumab was associated with a higher incidence of secondary bacterial infections including hospital acquired pneumonia and ventilator associated pneumonia (24 (50%) vs. 18 (28.6%). p=0.021). Additionally, there were three patients with fungal infections with one patient having the fungal infection in two different locations in the tocilizumab group compared with none in the non-tocilizumab group with statistical analysis showing a trend towards significance (p=0.078).

In multivariate logistic regression model to predict bacterial infections, the following independent variables were included: age, sex, APACHE II score, Charlson comorbidity index, use of immunosuppressive agents, presence of DVT/PE, diabetes with any organ damage and use of tocilizumab. Tocilizumab use was independently positively associated with the development of bacterial infections with an odds ratio of 2.83 (95% CI 1.14-7.39) p=0.0248. APACHE II score was also independently positively associated with bacterial infections (OR: 1.078, 95% CI: 1.01-1.16, p=0.021). History of DVT/PE was negatively associated with bacterial infections (OR: 0.106, 95% CI: 0.0008-0.939).

Due to well-known limitations with bacterial cultures in hospitalized and critically ill patients, we included both culture proven and suspected bacterial infections in our analysis. Both non-tocilizumab and tocilizumab groups had 10 culture proven bacterial infections. Eight patients in the non-tocilizumab and 14 patients in the tocilizumab group were treated for bacterial infection although the cultures were negative or non-diagnostic. There was no difference between two groups in terms of culture negative bacterial infections (p=0.368).

In addition to its effects on bacterial infections, the use of tocilizumab also affected outcomes. Compared to patients who did not receive tocilizumab, those who were prescribed tocilizumab had higher mortality (39.6% vs. 17.4%, p=0.016). In contrast, lesser number of patients were discharged home in the tocilizumab group (33.3% vs. 57.2%, tocilizumab and non-tocilizumab group, respectively) (Table 1).

Lastly, we performed post-mortem evaluation of 7 cases; 3 received tocilizumab and 4 did not. All three cases who received tocilizumab had evidence of pneumonia on pathology (Figure 1). Two of four patients who did not receive tocilizumab were nursing home residents with history of stroke and dementia and they died on the same day of admission. Their post-mortem evaluation showed evidence of aspiration pneumonia. The other two patients who did not receive tocilizumab were hospitalized for 4 and 12 days. Their lungs demonstrated only pathological changes consistent with diffuse alveolar damage without any evidence of pneumonia (Figure 1). These findings raise concerns about the use of tocilizumab in the presence of an infection to attenuate CRS. In particular, the occurrence of secondary bacterial infections may prolong ICU stays, and the occurrence of secondary fungal infections stands out as unusual in critical care patients without traditional risk factors (e.g. neutropenia).

![Figure 1.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/07/06/2020.05.15.20103531/F1.medium.gif)

[Figure 1.](http://medrxiv.org/content/early/2020/07/06/2020.05.15.20103531/F1)

Figure 1. Postmortem histopathology of lungs from COVID-19 patients.
Low (100x) and high power (200x) images of lungs from patients who died due to COVID-19. **A**. Organizing hyaline membranes are seen in the lung which has pre-existing emphysema (100x). Higher power shows fibrin, fibroblasts and mononuclear cells incorporated into the alveolar walls (200x). **B**. There is diffuse alveolar damage with hyaline membranes lining alveoli (100x). Higher power shows minimal inflammation with only a few mononuclear cells (200x). **C**. There is extensive intra-alveolar inflammation (neutrophils) in an otherwise normal lung (100x). On higher power, there is minimal alveolar wall thickening by inflammatory cells (also mainly neutrophils on myeloperoxidase staining and only rare lymphocytes) (200x). **D**. Majority of the sections from this case show organizing intra-alveolar fibrin (100x). Several foci of acute inflammation with alveolar filling are present, as seen here on higher power (200x).

## DISCUSSION

Receiving tocilizumab was associated with a significantly higher rate of secondary infections and mortality at our institution. Furthermore, review of the available autopsy findings suggest that bacterial pneumonia is not uncommon in patients who die from COVID-19, particularly if they have received tocilizumab previously. These findings raise concerns about the use of anti-IL-6 therapy to attenuate a cytokine-release-like syndrome in COVID-19.

Host response to the pathogen during sepsis is a double-edged sword. Infection is generally met by a desired host inflammatory response, which can at times be vigorous. For decades, physicians have speculated whether this host response is to blame, at least in part, for the multi-organ dysfunction associated with infection, as can be seen in sepsis. In fact, the most recent definition of sepsis includes the term “dysregulated host response to infection”. According to current thought, there is orchestration of temporally distinct phases of the immune response, consisting of both initial prompt upregulation and subsequent curtailment. Dysfunctional responses may lead to overwhelming infection or persistent inflammation. Cytokines must be viewed through this lens: IL-6 may contribute to organ injury and death, but it is also central to innate immunity and microbial clearance.

At this time, it remains difficult to characterize the stage of the inflammatory response on an individual basis in real time. Similarly, many characteristics seen in severe inflammation, including, within autoregulatory limits, hypotension, may in fact represent an adaptive response. A dramatic elevation in inflammatory cytokines or other markers such as ferritin may indicate a particularly brisk host response against a more severe infection. Without proper discriminatory values that reliably identify patients in whom hyper-inflammation is the key driver of pathogenesis, treatment with anti-inflammatory therapies may be detrimental. While there may be a subset of patients who may potentially benefit from the use of tocilizumab, current evidence does not support the routine use of tocilizumab or other drugs that regulate host immune response (i.e. anti-IL1, anti-TNFα) in COVID-19 or non-COVID-19 sepsis. Evidence of efficacy for IL-6 blockade currently exists in rheumatological diseases and to manage complications of T cell engaging therapies, which are driven by a primary immune response. To date, there is no convincing evidence that immune blockade is clinically beneficial when microbial infections drive the host response. Indeed, a previous anti-cytokine strategy targeting TNFα increased mortality in septic patients15 and was linked to increased risk of infections16.

There are several important limitations to our study. The retrospective nature and lack of randomization naturally do not allow for the establishment of causality. However, we would like to point out that the patients in both groups were relatively closely matched with regards to severity of illness (APACHE II scores of 15.76 (non-tocilizumab group) vs 17.38 (tocilizumab group), p=0.096). Patients who received tocilizumab may have developed more severe disease during the course of their ICU stay, however, our institutional policies regarding tocilizumab included important exclusion criteria, including LFT abnormalities, suspected ongoing infection, and enrollment in a clinical trial (e.g. remdesivir). In light of these limitations, it appears unlikely that the group receiving tocilizumab was excessively enriched with the sickest patients and gives credence to the descriptive power of the severity of illness captured by APACHE II scoring.

We did not restrict our definition to rely solely on microbiologic data but defined infection as: documented infection in the medical record and antibiotic therapy initiated and continued for >2 days. This definition reflects real-world circumstances, as microbiologic identification of the causative organism is frequently not achieved in infections in the ICU, in particular for hospital-acquired and ventilator-acquired pneumonias (HAP, VAP). Furthermore, the IDSA/ATS guideline definitions of HAP and VAP do not rely on microbiologic data. Nonetheless, it is possible that our definition is too broad and may capture some patients who were treated for clinical deterioration due to COVID-19 rather than a secondary infection. However, our definition is similar to other studies where diagnoses are based on coding. Restriction to only microbiologically proven cases would likely may miss a number of infections and may capture contaminants. While the numbers are small, the autopsy findings support the notion that bacterial pneumonia may be present even in absence of positive culture data.

Severe COVID-19 often elicits a strong inflammatory response with elevation in several cytokines, such as IL-6. This inflammatory phenotype shares clinical features with CAR-T cell-induced CRS and has led to the off-label use of tocilizumab for COVID-19. However, clinical similarity does not necessarily that changes seen in COVID-19 and CAR-T CRS are causally related and due to an overactive immune system. At least in part, the forceful immune response in COVID-19 may be adaptive and required for the anti-viral response. In our population of critically-ill COVID-19 patients, the use of tocilizumab was associated with an increase in infections and mortality. Our findings should raise physician awareness about the potential side effects of tocilizumab on pathogen clearance and development of secondary infections. It also calls for the urgent need to study all drugs including any immunosuppressive agents in randomized controlled trials to better understand their role in any hyperimmune response and on clearance of SARS-CoV-2 and other hospital-acquired pathogens.

## Data Availability

Data are available upon responsible request by contacting the corresponding author.

*   Received May 15, 2020.
*   Revision received July 6, 2020.
*   Accepted July 6, 2020.


*   © 2020, Posted by Cold Spring Harbor Laboratory

The copyright holder for this pre-print is the author. All rights reserved. The material may not be redistributed, re-used or adapted without the author's permission.

## REFERENCES

1.  1.Zhou F, Yu T, Du R, et al. Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study. Lancet 2020;395:1054–62.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/S0140-6736(20)30566-3&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F07%2F06%2F2020.05.15.20103531.atom) 

2.  2.Mehta P, McAuley DF, Brown M, et al. COVID-19: consider cytokine storm syndromes and immunosuppression. Lancet 2020;395:1033–4.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/S0140-6736(20)30628-0&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F07%2F06%2F2020.05.15.20103531.atom) 

3.  3.Kopf M, Baumann H, Freer G, et al. Impaired immune and acute-phase responses in interleukin-6-deficient mice. Nature 1994;368:339–42.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1038/368339a0&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=8127368&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F07%2F06%2F2020.05.15.20103531.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=A1994NB98500047&link_type=ISI) 

4.  4.Narazaki M, Kishimoto T. The Two-Faced Cytokine IL-6 in Host Defense and Diseases. Int J Mol Sci 2018;19.
    
    

5.  5.Lang VR, Englbrecht M, Rech J, et al. Risk of infections in rheumatoid arthritis patients treated with tocilizumab. Rheumatology (Oxford) 2012;51:852–7.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1093/rheumatology/ker223&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=21865281&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F07%2F06%2F2020.05.15.20103531.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000303159800014&link_type=ISI) 

6.  6.Pawar A, Desai RJ, Solomon DH, et al. Risk of serious infections in tocilizumab versus other biologic drugs in patients with rheumatoid arthritis: a multidatabase cohort study. Ann Rheum Dis 2019;78:456–64.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6MTE6ImFubnJoZXVtZGlzIjtzOjU6InJlc2lkIjtzOjg6Ijc4LzQvNDU2IjtzOjQ6ImF0b20iO3M6NTA6Ii9tZWRyeGl2L2Vhcmx5LzIwMjAvMDcvMDYvMjAyMC4wNS4xNS4yMDEwMzUzMS5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 

7.  7.Davis JS, Ferreira D, Paige E, Gedye C, Boyle M. Infectious Complications of Biological and Small Molecule Targeted Immunomodulatory Therapies. Clin Microbiol Rev 2020;33.
    
    

8.  8.Velazquez-Salinas L, Verdugo-Rodriguez A, Rodriguez LL, Borca MV. The Role of Interleukin 6 During Viral Infections. Front Microbiol 2019;10:1057.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.3389/fmicb.2019.01057&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F07%2F06%2F2020.05.15.20103531.atom) 

9.  9.Xu X, Han M, Li T, et al. Effective treatment of severe COVID-19 patients with tocilizumab. Proc Natl Acad Sci U S A 2020;117:10970–5.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NDoicG5hcyI7czo1OiJyZXNpZCI7czoxMjoiMTE3LzIwLzEwOTcwIjtzOjQ6ImF0b20iO3M6NTA6Ii9tZWRyeGl2L2Vhcmx5LzIwMjAvMDcvMDYvMjAyMC4wNS4xNS4yMDEwMzUzMS5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 

10. 10.Toniati P, Piva S, Cattalini M, et al. Tocilizumab for the treatment of severe COVID-19 pneumonia with hyperinflammatory syndrome and acute respiratory failure: A single center study of 100 patients in Brescia, Italy. Autoimmun Rev 2020;19:102568.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.autrev.2020.102568&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F07%2F06%2F2020.05.15.20103531.atom) 

11. 11.Morena V, Milazzo L, Oreni L, et al. Off-label use of tocilizumab for the treatment of SARS-CoV-2 pneumonia in Milan, Italy. Eur J Intern Med 2020;76:36–42.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F07%2F06%2F2020.05.15.20103531.atom) 

12. 12.Luo P, Liu Y, Qiu L, Liu X, Liu D, Li J. Tocilizumab treatment in COVID-19: A single center experience. J Med Virol 2020;92:814–8.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1002/jmv.25801&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F07%2F06%2F2020.05.15.20103531.atom) 

13. 13.Campochiaro C, Della-Torre E, Cavalli G, et al. Efficacy and safety of tocilizumab in severe COVID-19 patients: a single-centre retrospective cohort study. Eur J Intern Med 2020;76:43–9.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.ejim.2020.05.021&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=http://www.n&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F07%2F06%2F2020.05.15.20103531.atom) 

14. 14.Mansournia MA, Geroldinger A, Greenland S, Heinze G. Separation in Logistic Regression: Causes, Consequences, and Control. Am J Epidemiol 2018;187:864–70.
    
    

15. 15.Fisher CJ, Jr.., Agosti JM, Opal SM, et al. Treatment of septic shock with the tumor necrosis factor receptor:Fc fusion protein. The Soluble TNF Receptor Sepsis Study Group. N Engl J Med 1996;334:1697–702.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1056/NEJM199606273342603&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=8637514&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F07%2F06%2F2020.05.15.20103531.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=A1996UT40000003&link_type=ISI) 

16. 16.Ali T, Kaitha S, Mahmood S, Ftesi A, Stone J, Bronze MS. Clinical use of anti-TNF therapy and increased risk of infections. Drug Healthc Patient Saf 2013;5:79–99.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.2147/DHPS.S28801&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=23569399&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F07%2F06%2F2020.05.15.20103531.atom)