Immunosuppressive Drugs and COVID-19: A Review

Background Coronavirus disease 2019 (COVID-19) is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). It is currently unknown whether immunosuppressive drugs are advantageous or detrimental in patients with COVID-19. Immunosuppressive drugs could be harmful in the initial phase of COVID-19. In this phase, the host immune response is necessary to inhibit viral replication. However, immunosuppressive drugs might have a beneficial effect in the later, more severe phase of COVID-19. In this phase, an overshoot of the host immune response (the “cytokine storm”) can cause ARDS, multiorgan failure and mortality. Aim To summarize the available evidence on the effect of immunosuppressive drugs on infection with SARS-CoV-2. The effects of immunosuppressive drugs on similar pandemic coronaviruses may resemble the effects on SARS-CoV-2. Thus, we also included studies on the severe acute respiratory syndrome coronavirus (SARS-CoV) and Middle East respiratory syndrome coronavirus (MERS-CoV) Methods The study protocol was registered in PROSPERO (registration number CRD42020181137). We included randomized controlled trials (RCTs), cohort studies with a control group and case-control studies concerning humans ≥ 18 years old. We also included in-vitro studies and animal studies with a control group. Results and Conclusion Sixty-nine studies were included. Interestingly, MPA inhibits SARS-CoV-2 replication in-vitro. Clinical studies are needed to confirm the inhibitory effect of MPA on SARS-CoV-2 replication in-vivo. There are indications that corticosteroids and IL-6 inhibitors, like tocilizumab, can reduce mortality and prevent mechanical ventilation in patients with COVID-19. However, observational studies have contradictory results and the risk of bias is high. Thus, these results have to be confirmed in high-quality clinical trials before these drugs can be implemented as standard care. Based on the positive results of CNIs, mTOR inhibitors and thiopurine analogues in in-vitro studies with SARS-CoV and MERS-CoV, it would be interesting to investigate their effects on SARS-CoV-2 replication.


INTRODUCTION
Coronavirus disease 2019 , caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has a mild course in the majority of patients. However, approximately 14% of the patients require hospitalization for oxygen support and 5% is admitted to the intensive care unit (ICU) (World Health Organization (WHO), 2020).
COVID-19 has a triphasic course (Siddiqi and Mehra, 2020). In the first phase, patients have mild respiratory and systemic symptoms such as dry cough, malaise and fever. An adequate response of the innate and adaptive immune system can eliminate the virus and preclude disease progression to the next, more severe stages (Shi et al., 2020). In the second phase, viral multiplication and localized inflammation of the lung tissue occurs, causing viral pneumonia. A minority of patients with COVID-19 will undergo a transition into the third and most severe phase of illness: a syndrome of systemic hyperinflammation, also referred to as secondary hemophagocytic lymphohistiocytosis or the cytokine storm syndrome. These patients have high levels of proinflammatory cytokines, such as interleukin (IL)-2, IL-6, IL-7 and tumor-necrosis factor-a (TNF-a) Zhang et al., 2020). In this phase, patients can develop acute respiratory distress syndrome (ARDS) and multiorgan failure, which are the main causes of mortality of COVID-19 (Cao et al., 2020;Du et al., 2020;Mehta et al., 2020;Ruan et al., 2020;Wu et al., 2020).
It is hypothesized that immunosuppressive drugs could be used to prevent and treat the hyperinflammatory phase of COVID-19. However, these drugs inhibit the host immune response against the virus as well. Thus, they could be harmful in earlier stages of COVID-19. The net effect of immunosuppressive drugs in patients with COVID-19 is currently subject of debate.
The primary aim of this review is to summarize the available evidence on the effect of immunosuppressive drugs on infection with SARS-CoV-2. Given the short time period since the start of the COVID 19 pandemic, relatively few studies on COVID-19 are completed. Therefore, we also searched for papers reporting the effect of immunosuppressive drugs on infection with severe acute respiratory syndrome coronavirus (SARS-CoV) and Middle East respiratory syndrome coronavirus (MERS-CoV), two other pandemic coronaviruses belonging to the Nidovirales order (de Wit et al., 2016;Huang et al., 2020). These viruses are the causative agents of severe acute respiratory syndrome (SARS) and Middle East respiratory syndrome (MERS), which share many features with COVID-19 (Wang et al., 2004;Yin and Wunderink, 2018;Wang D. et al., 2020). Thus, the effects of immunosuppressive drugs on these viruses may resemble the effects on SARS-CoV-2. The results of this review can be used in the development of evidence-based guidelines for the management of patients with COVID-19 and of patients using immunosuppressive drugs during the COVID-19 pandemic.

METHODS
We conducted a PubMed search, without limitations for the publication date or publication type. In the search, the terms "COVID-19", "SARS", "MERS" (and multiple synonyms) were combined with search terms for the different classes of immunosuppressive drugs (see: Supplementary File 1). The last search was performed on 8 July 2020. All papers were screened by title and abstract, and the full text of potentially eligible papers was read. In addition, the reference list of all identified papers was screened manually. To be included in our review, studies had to meet the following inclusion criteria: 1. Studies concerning humans ≥ 18 years old, animals or cells infected with SARS-CoV-2, SARS-CoV or MERS-CoV. 2. In-vitro or in-vivo treatment with any of the following immunosuppressive drugs: calcineurin inhibitors (CNIs; cyclosporine (CsA), tacrolimus(TAC)), antimetabolites (like mycophenolic acid (MPA), azathioprine (AZA), methotrexate), mammalian target of rapamycin (mTOR) inhibitors (sirolimus, everolimus), corticosteroids (like methylprednisolone, hydrocortisone, prednisone, dexamethasone), cyclophosphamide, rituximab, alemtuzumab, IL-6 inhibitors (like tocilizumab), basiliximab, anakinra, dupilumab, brodalumab, secukinumab, ixekizumab, anti-TNF-a inhibitors (like infliximab), abatacept, belatacept, or eculizumab. 3. Data on one or more of the following outcome measures: viral load, viral replication, clinical outcome (e.g. mortality rate, ICU admission rate, length of hospital stay). 4. Study type: in-vitro study, animal study with control group, randomized controlled trial (RCT), cohort study with control group, case-control study. 5. Language: English.
We categorized the obtained information per immunosuppressive drug class. In addition, we sorted the data according to the type of outcome parameter, i.e. viral load or clinical outcome. The study protocol was registered in PROSPERO (registration number CRD42020181137).

RESULTS
The database search yielded 1939 search hits, with 69 studies matching the inclusion criteria. A summary of the results is presented in Table 1.

Corticosteroids
Corticosteroids have a wide range of anti-inflammatory and immunomodulatory effects, including inhibition of the synthesis of pro-inflammatory cytokines, reduction of leucocyte trafficking, and induction of apoptosis of T-lymphocytes (Lansbury et al., 2019). In addition, corticosteroids might increase the sensitivity to vasopressors (Lansbury et al., 2019) and can be used to treat adrenal insufficiency, which is present in 20% of the patients with critical illness, and in up to 60% of those with sepsis and septic shock .
In Table 2, the 38 clinical studies matching the inclusion criteria are summarized.

Invitro
Animal Human Viral load and viral replication Clinical outcome Corticosteroids SARS-CoV-2 0 0 18 Cohort studies have different results. 5 cohort studies found that steroid use had no effect on SARS-CoV-2 clearance time. 3 cohort studies report that steroid use associated with longer SARS-CoV-2 clearance time. All studies have a high risk of confounding (by indication).
One RCT found that dexamethasone use was associated with lower 28-day mortality rate, shorter length of hospital stay and lower prevalence of mechanical ventilation. Limitations: open-label. Observational studies have different results: 6 cohort studies found that steroids have a beneficial effect, 2 cohort studies conclude that steroids have a detrimental effect and 1 cohort studies report that steroids have no effect on mortality. The risk of confounding (by indication) is high in all studies. 3 cohort studies report that higher steroid dose is associated with death, but they have a significant risk of confounding by indication. SARS-CoV 0 0 18 One RCT found that steroid use had no effect on SARS-CoV clearance time, but the risk of confounding was high. One autopsy study found that steroid use was not associated with viral load in lung tissue, but this study may have been underpowered.
Cohort studies have different results: 2 cohort studies found that steroids have a beneficial effect, 7 cohort studies conclude that steroids have a detrimental effect and 3 cohort studies report that steroids have no effect on mortality. Moreover, the risk of confounding (by indication) is high in all studies. Neither can we draw conclusions on the ideal timing of steroid administration or the ideal steroid dose. Cohort studies report conflicting results and have high risk of confounding (by indication). MERS-CoV 0 0 2 One cohort study in ICU patients found that steroid use was associated with longer MERS-CoV clearance time. There is a high risk of confounding (by indication).
One cohort study found that steroid use was associated with higher mortality. The risk of bias could not be assessed because of incomplete baseline characteristics. In 1 cohort study in ICU patients had no clear conclusion, because several statistical methods provided different results. One cohort study found that MMF use was associated with a lower mortality rate, but there is significant risk of confounding by indication.

Viral Load
1. SARS-CoV-2: Five observational studies found that SARS-CoV-2 clearance time did not differ between steroid-treated patients and those not treated with steroids Xu et al., 2020;Yuan et al., 2020;Zha et al., 2020;Zheng et al., 2020). Contrarily, three other observational studies report that steroid treatment was associated with a longer time until SARS-CoV-2 clearance Ling et al., 2020;Xu et al., 2020). The risk of confounding was high in all these studies. 2. SARS-CoV: In a double-blinded RCT, patients with SARS were randomized to treatment with intravenous hydrocortisone (300 mg/day for 12 days) or placebo to study the effect of hydrocortisone on SARS-CoV clearance time . Patients with comorbidity, respiratory failure or symptoms for ≥ 5 days were excluded. In the second and third week of illness, the mean SARS-CoV RNA plasma levels were higher in the hydrocortisone group than in the placebo group, but the SARS-CoV clearance time did not differ between both groups. However, the risk of confounding is high, because high-dose methylprednisolone (500 mg/day) was given to 5 out of 10 patients in the hydrocortisone group and to 6 out of 7 patients in the placebo group. In a retrospective cohort study, steroid use was not associated with the viral load in lung tissue obtained at autopsy in 11 SARS patients (Mazzulli et al., 2004). However, the study may have been underpowered and the risk of confounding is high.
3. MERS-CoV: In an observational study in patients with MERS admitted to the ICU, steroid use was associated with a delay in MERS-CoV RNA clearance (Arabi et al., 2018). This study has a high risk of confounding.

SARS-CoV-2:
In an open-label RCT, patients with COVID-19 were randomized in a ratio of 1:2 to treatment with dexamethasone (6 mg daily, oral or intravenous, for a maximum of 10 days) at hospital admission or standard care (controls) (Recovery Collaborative Group et al., 2020). Age, sex, comorbidities and the requirement for respiratory support were similar in both groups. The median number of days since onset of symptoms was 8 days (IQR 5-13 days) for patients in the dexamethasone group and 9 days (IQR 5-13 days) for controls. Overall, the 28-day mortality rate was lower in dexamethasone-treated patients compared to controls (21.6 vs. 24.6%; age-adjusted RR 0.83, 95% CI 0.75-0.93). Similarly, dexamethasone use was associated with a lower 28-day mortality rate in the subgroups of patients requiring oxygen (RR 0.82, 95% CI 0.72-0.94) or invasive mechanical ventilation (RR 0.64, 95% CI 0.51-0.81) at randomization. Only in the subgroup of patients who were not treated with any type of respiratory support at randomization, there was no significant difference in the 28-day mortality rate between dexamethasone-treated patients and controls (RR 1.19, 95% CI 0.91-1.55).   Methyl-pred 1-2 mg/kg/day In patients treated with methylpred, mechanical ventilation was less prevalent, length of ICU and total hospital stay were shorter, and CRP and IL-6 decreased faster than in patients not treated with steroids. The mortality rate was not significantly different.

Risk of confounding (by indication).
(Continued) Methyl-pred 0.5-1 mg/kg/day The rate of SARS-CoV-2 clearance and SARS-CoV-2 antibody formation did not differ between steroidtreated and steroid-free patients.
Recovery of radiologic images was not different in steroidtreated patients as compared to steroid-free patients.   The time until SARS-CoV was no longer detectable in plasma did not differ between patients in the hydrocort group and controls. The mean SARS-CoV RNA plasma level in the second and third week of illness were higher in the hydrocort than placebo group.

MERS
One patient was admitted to the ICU and eventually died; all other patients recovered and were discharged.

Risk of confounding.
High-dose methylpred was given to 6 out of 7 patients in the placebo group, and 5 out of 10 patients in the hydrocort group.
Dexamethasone use was also associated with a lower rate of invasive mechanical ventilation (RR 0.76, 95% CI 0.61-0.96) and shorter length of hospital stay (median 12 vs. 13 days). Six retrospective cohort studies also found that the use of steroids was associated with beneficial outcomes in patients with COVID-19 (Fadel et al., 2020;Fernandez Cruz et al., 2020;Lu et al., 2020;Sanz Herrero et al., 2020;Wang Y. et al., 2020;Wu et al., 2020). Two of these studies found that patients treated with steroids had a lower prevalence of mechanical ventilation and shorter length of ICU and hospital stay than controls (Fadel et al., 2020;. In three other studies that only included COVID-19 patients with ARDS, the mortality rate was lower in patients treated with steroids than in controls (Fernandez Cruz et al., 2020;Lu et al., 2020;Wu et al., 2020). In COVID-19 patients treated with tocilizumab, additional administration of steroids was also associated with a lower mortality rate (Sanz Herrero et al., 2020).
In contrast, two retrospective cohort studies found that steroid use was associated with detrimental outcome (Giacobbe et al., 2020;Mo et al., 2020). The first study included COVID-19 patients who were admitted to the ICU. They found that methylprednisolone use was independently associated with development of a blood stream infection (Giacobbe et al., 2020). Another retrospective cohort study found that patients treated with steroids were less likely to reach clinical and radiological remission within 10 days than patients who did not use steroids (Mo et al., 2020). However, this study has a high risk of confounding by indication.
Lastly, one cohort study in hospitalized COVID-19 patients found that the rate of steroid treatment was not significantly different between surviving and deceased patients (Cao et al., 2020). Again, there was significant risk of confounding. 2. SARS-CoV: Similar to the results for SARS-CoV-2, observational studies in patients with SARS have different results. In one cohort study that included patients with severe SARS, steroid use was associated with reduced mortality (Chen et al., 2006). Three other cohort studies found that treatment with steroids had no effect on the mortality rate (Chen et al., 2006;Wei et al., 2009) or development of ARDS (Chen et al., 2005) in patients with SARS. Several other cohort studies report that steroid use was associated with detrimental outcome (Auyeung et al., 2005;Leung et al., 2005;Ng et al., 2004;Wang et al., 2005;Xie et al., 2005;Yam et al., 2007). Two retrospective cohort studies report that steroid use was associated with a higher risk of ICU admission and mortality (Auyeung et al., 2005;Yam et al., 2007). In two studies that included patients who recovered from SARS, steroid use was associated with the presence of CT abnormalities and a lower diffusion capacity of the lungs during follow-up Wang et al., 2005;Xie et al., 2005). However, the risk of confounding by indication is high. Another cohort study investigated autopsy findings of 8 SARS patients with ARDS. They found that all patients with hydrocortisone had myofiber atrophy, whereas this was absent in patients who were not treated with steroids (Leung et al., 2005). However, critical illness neuropathy could not be distinguished from steroid myopathy in this study. 3. MERS-CoV: One retrospective cohort study included MERS patients who were admitted to the ICU. Steroid use was associated with a higher 90-day mortality in a multivariable logistic regression model. However, in a Cox proportional hazard model, steroid use was not associated with a higher mortality rate (Arabi et al., 2018). This suggests that their statistical models did not adjust for all confounders. Another retrospective cohort study found that steroid use was associated with a higher mortality in hospitalized MERS patients (Alfaraj et al., 2019). It was not possible to assess the risk of bias in this study, because essential information is missing, such as baseline characteristics.

Effect of Steroid Dose
1. SARS-CoV-2: Two retrospective cohort studies found that a higher steroid dose was associated with death and prolonged time until viral clearance , while use of lowdose steroids was not . Another retrospective cohort study found that each 10 mg increase in steroid dose was associated with a 4% increase in the risk of mortality after 28 days in COVID-19 patients with ARDS or sepsis . However, the steroid dose tended to be the highest in the patients who were most severely ill. Thus, there is significant risk of confounding by indication in these studies. 2. SARS-CoV: In one retrospective cohort study, a lower mean steroid dose was associated with a higher 28-day mortality and mechanical ventilation rate (Gomersall et al., 2004). Contrarily, another retrospective cohort study found that a higher mean steroid dose was associated with a higher mortality rate (Wei et al., 2009). Two other retrospective cohort studies found no association between steroid dose and clinical outcome (Ho et al., 2003;Hui et al., 2005). The first study found that the rate of ICU admission, mechanical ventilation or mortality after 3 weeks were not significantly different for patients treated with high-dose or low-dose steroids (Ho et al., 2003). The other study reports that steroid dose was not associated with the results of the 6minute walking test or the severity of radiographic abnormalities at 6 months after SARS . In all of these studies, the risk of confounding (by indication) is high. 3. MERS-CoV: No studies available.

Timing of Steroid Administration
1. SARS-CoV-2: No studies available. 2. SARS-CoV: Two observational studies investigated whether the timing of steroid administration was associated with clinical outcomes in patients with SARS (Zhao et al., 2003;Jang et al., 2004). The first study reports that the time between onset of symptoms and start of steroids was not associated with the risk of ICU admission or mechanical ventilation (Jang et al., 2004). The second study found that early treatment with high-dose methylprednisolone might be beneficial for SARS patients (Zhao et al., 2003). In this study, treatment with methylprednisolone was started at different moments in the course of the disease: (A) if patients had not recovered after 14 days, (B) if symptoms or radiological abnormalities worsened, or (C) if fever persisted for ≥ 3 days after admission. The rate of mechanical ventilation was 7, 33, and 0%, respectively, and the mortality rate was 6, 12, and 0%, respectively. However, patients were not randomized to different treatment protocols, and consequently, baseline characteristics and disease severity were different. In both studies, the risk of confounding is high.

Calcineurin Inhibitors (CNI): Cyclosporin A (CsA) and Tacrolimus (TAC)
CNIs (cyclosporine A (CsA) and tacrolimus (TAC)) are used to prevent rejection after organ transplantation and to treat autoimmune diseases, like inflammatory bowel disease. CsA and TAC inhibit T-cell activation. In-vivo, CsA and TAC form complexes with cyclophilins and FK506-binding proteins, respectively. These complexes prevent the phosphatase activity of calcineurin. As a result, the dephosphorylation of the nuclear factor of activated T cells is decreased (Carbajo-Lozoya et al., 2012;Ma et al., 2016). In addition, cyclophilins, the binding proteins of CsA, catalyze the cis/trans isomerization of propyl peptide bonds. This is an essential step in correct folding of proteins, such as cellular and viral proteins (Ma-Lauer et al., 2020). This function of cyclophilin A is found to be essential for the replication of SARS-CoV-2 and other viruses belonging to the Nidovirales order (Carbajo-Lozoya et al., 2014).  Pfefferle et al., 2011;Carbajo-Lozoya et al., 2012;de Wilde et al., 2013;Li et al., 2018;Sauerhering et al., 2020). One of these studies found that a high concentration of CsA (15 µM) completely inhibited the CPE, without affecting the viability of the cells (de Wilde et al., 2013). Next to these in-vitro effects, CsA also inhibited MERS-CoV viral replication and reduced cellular apoptosis in ex-vivo cultures of bronchial and lung tissue (Li et al., 2018).

Clinical Outcome
No studies matching the inclusion criteria.

Mycophenolic Acid (MPA)
Mycophenolic acid (MPA) and its prodrugs, mycophenolate mofetil (MMF) and mycophenolate sodium, are used in the treatment of autoimmune diseases and to prevent rejection in organ transplant recipients. MPA inhibits inosine-5'monophosphate dehydrogenase, which leads to depletion of intracellular guanosine and deoxyguanosine nucleotides. This suppresses DNA synthesis and thus proliferation of T and B lymphocytes (Villarroel et al., 2009).

Viral Load
1. SARS-CoV-2: One in-vitro study found that MPA inhibits SARS-CoV-2 replication in VeroE6/TMPRSS2 cells (Kato et al., 2020). In another study (Han et al., 2020), human pluripotent stem cells (hPSC) were differentiated into lung organoids and then infected with SARS-CoV-2. In these lung organoids, MPA inhibited viral replication while the CPE of SARS-CoV-2 was still observed, even with high concentrations of MPA. 2. SARS-CoV: MPA does not inhibit the proteolytic activity of SARS-CoV PL pro (Cheng et al., 2015) or SARS-CoV replication in Vero cells (Barnard et al., 2006). 3. MERS-CoV: Two studies showed that MPA effectively inhibits the proteolytic activity of the papain-like protease (PL pro ) of MERS-CoV (Cheng et al., 2015;Lin et al., 2018). PL pro is responsible for the cleavage of nonstructural proteins, which are essential for viral maturation. Three other in-vitro studies showed that MPA significantly inhibited the replication and CPE of MERS-CoV in Vero cells (Chan et al., 2013;Hart et al., 2014;Shen et al., 2019). This effect was dose-dependent.
In contrast, an in-vivo study in marmosets infected with MERS-CoV found that the mean viral load in the lungs was higher in MMF-treated animals than in controls (Chan et al., 2015). However, since MERS-CoV does not cause lethal disease in marmosets, this animal model does not adequately resemble human MERS (Johnson et al., 2015). In univariate analysis, MMF treatment was associated with survival. However, MMF was given to the less severely ill patients and, seven of the eight MMF-treated patients (87.5%) were also treated with interferon beta (Al Ghamdi et al., 2016). Thus, there is significant risk of confounding.

Thiopurine Analogues
Thiopurine analogues (azathioprine (AZA) and 6-mercaptopurine (6MP)) are used as anticancer treatment, to prevent rejection in organ transplant recipients and as treatment of several chronic autoimmune diseases. AZA is a prodrug of 6MP. In-vivo, 6MP is converted into 6-thioguanine (6TG) which is incorporated into cellular DNA. This prevents further DNA replication (Chen et al., 2009).

Viral Load
1. SARS-CoV-2: No studies available. 2. SARS-CoV and MERS-CoV: 6MP and 6TG effectively inhibit the proteolytic activity of PL pro of MERS-CoV and SARS-CoV in a dose-dependent manner (IC 50 of 12 to 27 µM) in inhibition assays with peptide and fluorogenic substrates (Chou et al., 2008;Cheng et al., 2015;Lin et al., 2018). These results suggest that 6MP and 6TG can inhibit the replication of MERS-CoV and SARS-CoV in-vitro.

Clinical Outcome
No studies matching the inclusion criteria.

Clinical Outcome
No studies matching the inclusion criteria.

Anti-Cytokine Agents
Anti-Tumor-Necrosis-Factor-a (TNF-a) Agents There are several types of anti-tumor-necrosis-factor-a (TNF-a) agents: infliximab, adalimumab and golimumab (monoclonal antibodies against TNF-a), certolizumab (TNF-a binding fragment of a monoclonal antibody) and etanercept (fusion protein composed of the extracellular portion of the TNFreceptor-2 and the Fc portion of immunoglobulin G1). Anti-TNF-a agents are used in the treatment of autoimmune diseases, like rheumatoid arthritis and inflammatory bowel disease. TNFa is a pro-inflammatory cytokine that recruits neutrophils and monocytes to the area of inflammation and activates intracellular signaling in various cells of the immune system (Mitoma et al., 2018).

Viral Replication
No studies matching the inclusion criteria.

Clinical Outcome
1. SARS-CoV-2: No studies available. 2. SARS-CoV: In an animal study, mice infected with SARS-CoV were treated with an anti-TNF-a monoclonal antibody or an isotype-matched control antibody. In mice treated with the anti-TNF-a monoclonal antibody, the onset of weight loss and respiratory illness was delayed compared to controls. However, the mortality rate after 10 days was similar in both groups (Nagata et al., 2008). 3. MERS-CoV: No studies available.

Anakinra
Anakinra is an IL-1 receptor antagonist that is registered for the treatment of several autoinflammatory diseases, such as adultonset Stills disease and familial Mediterranean fever .

Viral Replication
No studies matching the inclusion criteria.

Clinical Outcome
1. SARS-CoV-2: A retrospective cohort study  included patients with COVID-19 with moderate-tosevere ARDS and hyperinflammation (CRP ≥ 100 mg/L and/ or ferritin ≥ 900 ng/mL). Patients treated with mechanical ventilation or other anti-inflammatory agents were excluded. Patients in the intervention group (n = 29) received highdose anakinra (5mg/kg twice daily intravenously). The control group (n = 16) consisted of patients who retrospectively met the eligibility criteria for anakinra, but who presented to the hospital before the availability of the drug. The 3-week mortality rate was lower in anakinratreated patients compared to controls (HR = 0.20, 95% CI 0.04-0.63, p = 0.009), but more anakinra-treated patients than controls still required mechanical ventilation after 3 weeks (17 vs. 6%). The percentage of patients that was discharged and had resumed normal activities after 3 weeks was not significantly different (45 vs. 44%). A limitation of this study is the significant risk of confounding. For example, there were significant differences between patients treated with anakinra and controls. Tocilizumab and other IL-6 Inhibitors IL-6 is an important pro-inflammatory cytokine that is involved in the acute phase response and differentiation and function of B and T cells (Mosharmovahed et al., 2020). Of the three commercially available IL-6 inhibitors (tocilizumab, sarilumab and siltuximab), tocilizumab is the most well-known (Khiali et al., 2020). Tocilizumab is a recombinant humanized anti-IL-6 receptor monoclonal antibody. Tocilizumab is used to treat several autoinflammatory diseases, like rheumatoid arthritis, giant cell arteritis and the chimeric antigen receptor T-cell induced cytokine storm syndrome (Khiali et al., 2020). Notably, Il-6 inhibitors were not available at the time of the SARS and MERS pandemics. The nine retrospective cohort studies matching the inclusion criteria are summarized in Table 3.

Viral Replication
No studies matching the inclusion criteria.

Clinical Outcome
1. SARS-Cov-2: The nine studies matching the inclusion criteria report conflicting results. One retrospective cohort study found that treatment of COVID-19 patients with tocilizumab was associated with a higher mortality rate and higher prevalence of bacterial infection (Quartuccio et al., 2020). However, tocilizumab-treated patients were more severely ill than controls. They were also significantly older and more frequently received antivirals and glucocorticoids than controls.
In four other studies, the mortality and ICU admission rate were not significantly different between patients treated with an IL-6-inhibitor (tocilizumab or sarilumab) and controls Colaneri et al., 2020;Della-Torre et al., 2020;Rojas-Marte et al., 2020). Similarly, the discharge rate , duration of hospitalization , blood stream infections (Giacobbe et al., 2020), serious adverse event rate (including infections and bacteremia)  and rate of clinical improvement after 28 days  were not significantly different. In three of these studies (Colaneri et al., 2020;Della-Torre et al., 2020;Rojas-Marte et al., 2020), the controls were matched to the tocilizumab-treated patients on several important patient characteristics in order to reduce the risk of confounding by indication. In the fourth study , the patient characteristics and disease severity at admission were not significantly different between both groups and the eligibility criteria for tocilizumab were clearly defined. Moreover, controls were recruited among patients who fulfilled these eligibility criteria before and after the time period of tocilizumab availability.
Lastly, three other studies report a higher 3-week survival rate (Capra et al., 2020), a lower rate of mechanical ventilation (Klopfenstein et al., 2020) and a lower rate of ICU admission (Moreno-Garcıá et al., 2020) in patients treated with tocilizumab. Remarkably, in two of these studies, the tocilizumab-treated patients were more critically ill at admission than controls, but their outcome was better.
In summary, one study found that tocilizumab was associated with a higher mortality rate, but this study has a high risk of confounding. Four observational studies found no effect of tocilizumab or sarilumab and four other studies showed a beneficial effect of tocilizumab.  In multivariate analysis, tocilizumab use was significantly associated with better outcome (OR for ICU admission and/or death 0.03, 95% CI 0.007 -0.1).
Risk of confounding: moderate (tocilizumab-treated patients were more severely ill, had higher CRP and were more frequently treated with steroids than controls; however, in case of bias this would result in a more detrimental outcome in tocilizumab-treated patients (which was not observed)).

DISCUSSION
There are 30 studies meeting the inclusion criteria for our review that studied the effect of immunosuppressive drugs on infection with SARS-CoV-2. These studies investigated the effects of corticosteroids, mycophenolic acid (MPA), anakinra or tocilizumab. Interestingly, MPA inhibits SARS-CoV-2 replication in-vitro (Han et al., 2020;Kato et al., 2020). This is in contrast with COVID-19 guidelines aimed at patients using immunosuppressive drugs for a pre-existing disease British Transplantation Society, 2020; SocietéFrancophone de Transplantation (SFT) SFdN et al., 2020). These expert-opinionbased guidelines recommend to discontinue MPA or reduce the MPA dose in patients with COVID-19. Based on the results of our review, this approach can be questioned. Clinical studies are needed to confirm the inhibitory effect of MPA on SARS-CoV-2 replication in-vivo.
The results of some clinical studies suggest that corticosteroids are beneficial in patients with COVID-19, especially in mitigating the effects of the cytokine storm. An RCT found that dexamethasone use was associated with a lower 28-day mortality rate, a shorter length of hospital stay and a lower prevalence of mechanical ventilation (Recovery Collaborative Group et al., 2020). The beneficial effect of dexamethasone was greatest in the most severely ill patients. For example, in patients treated with mechanical ventilation, dexamethasone reduced the risk of mortality with 36% (95% CI 19-49%), whereas dexamethasone had no effect on the mortality rate of patients who did not require oxygen support. The results of this RCT support our hypothesis that immunosuppressive drugs can be used to prevent and treat the hyperinflammatory phase of COVID-19.
Six cohort studies confirmed that steroid use was associated with a better clinical outcome in patients with COVID-19 (Fadel et al., 2020;Fernandez Cruz et al., 2020;Lu et al., 2020;Sanz Herrero et al., 2020;Wu et al., 2020). However, two cohort studies have opposite results (Giacobbe et al., 2020;Mo et al., 2020). They found that steroid use was associated with a higher risk of blood stream infections (Giacobbe et al., 2020) and a lower 10-day remission rate (Mo et al., 2020). Unfortunately, all cohort studies have a high risk of confounding (by indication), which limits their validity and generalizability. Thus, more high-quality studies are needed to confirm the effect of steroids on the clinical outcome of COVID-19.
Moreover, it has to be established whether corticosteroids have an effect on the host immune response against the virus. An adequate response of the immune system in the first phase of COVID-19 can prevent progression to viral pneumonia (Shi et al., 2020). As expected, there were no studies that found that corticosteroid use was associated with accelerated SARS-CoV-2 clearance. However, some cohort studies found that steroid use was associated with a delay in SARS-CoV-2 clearance Ling et al., 2020;Xu et al., 2020), while other studies report that steroid use was not associated with SARS-CoV-2 clearance time Xu et al., 2020;Yuan et al., 2020;Zha et al., 2020;Zheng et al., 2020). These studies have a high risk of confounding (by indication). Thus, we could not draw firm conclusions on the effect of corticosteroids on viral replication.
There is growing interest in IL-6 inhibitors, like tocilizumab. Patients with severe COVID-19 have high levels of IL-6 ( Mosharmovahed et al., 2020) and in three cohort studies, tocilizumab use was associated with lower mortality and ICU admission rate (Capra et al., 2020;Klopfenstein et al., 2020;Moreno-Garcıá et al., 2020). However, other cohort studies report contradictory results Colaneri et al., 2020;Della-Torre et al., 2020;Giacobbe et al., 2020;Quartuccio et al., 2020;Rojas-Marte et al., 2020). Furthermore, the risk of confounding in these observational studies is high. As a result, it is not clear yet whether IL-6 inhibitors can be used to prevent and treat the cytokine storm. Fortunately, several research groups have published a protocol for a RCT to assess the efficacy and safety of tocilizumab as treatment of COVID-19 (Cellina et al., 2020;Fu et al., 2020;Rilinger et al., 2020).
The observational studies for corticosteroids and IL-6 inhibitors report contradictory results, which is probably related to heterogeneity in study populations. In addition, there could have been differences in local treatment protocols. These factors limit the possibility of conducting a meta-analysis. Furthermore, in many studies there are differences in patient characteristics (such as age, gender and comorbidity) and treatment regimens between patients in the intervention group and control group. Moreover, in many cases, the most severely ill patients were treated with immunosuppressive drugs, whereas less severely ill patients were not. This might have resulted in confounding by indication.
Currently, the number of studies on the effect of immunosuppressive drugs on COVID-19 is still limited. SARS and MERS are also caused by a coronavirus and share many features with COVID-19 (Wang D. et al., 2004;Yin and Wunderink, 2018;Wang D. et al., 2020). Thus, the effects of immunosuppressive drugs on these viruses may resemble the effects on SARS-CoV-2. We were able to include 39 additional studies that studied the effect of immunosuppressive drugs on infection with SARS-CoV and MERS-CoV. CNIs and mTOR inhibitors inhibit the viral replication of SARS-CoV and MERS-CoV in-vitro. There is also indirect evidence that thiopurine analogues inhibit the viral replication of SARS-CoV. It would therefore be interesting to investigate the effect of CNIs, mTOR inhibitors and thiopurine analogues on the viral replication of COVID-19.
Notably, there are some important differences between SARS-CoV-2, SARS-CoV and MERS-CoV. First, the nucleotide sequence of the genome of SARS-CoV-2 is only 79.7% similar to SARS-CoV . However, the similarity of the nucleotide sequence of envelope, nucleocapsid and spike proteins is as high as 96, 90, and 77%, respectively . The similarity of the genome of SARS-CoV-2 and MERS-CoV is only 54%, and for the spike protein, a transmembrane protein involved in binding of the virus to the host receptor, 31.9% . Second, MERS-CoV uses dipeptidyl peptidase 4 (DPP4) as host cell receptor , while SARS-CoV-2 and SARS-CoV use ACE2 (Li et al., 2003;Hoffmann et al., 2020). Lastly, the case fatality rate of MERS-CoV is much higher than of COVID-19 and SARS. While interpreting the results of this review, these differences should be kept in mind.
Our study has some limitations. First, we did not fully meet all the requirements of a systematic review, since we have only searched one database and there was no second reviewer. Second, we only included "genuine" immunosuppressive drugs in our review. As a consequence, we excluded antiviral/antiparasitic drugs, like (hydroxy)chloroquine, despite the fact that some would argue that this drug has immunosuppressive properties.

CONCLUSIONS
Some immunosuppressive drugs may be beneficial in the treatment of COVID-19. MPA inhibits SARS-CoV-2 replication in-vitro. There are indications that corticosteroids and IL-6 inhibitors, like tocilizumab, can reduce mortality and prevent mechanical ventilation in patients with COVID-19. These results have to be confirmed in high-quality clinical trials before these drugs can be implemented as standard care. Based on the positive results of CNIs, mTOR inhibitors and thiopurine analogues in in-vitro studies with SARS-CoV and MERS-CoV, it would be interesting to investigate their effects on SARS-CoV-2 replication.

AUTHOR CONTRIBUTIONS
RM: idea for the review. All authors: research design. TS: PubMed search, data extraction, writing first draft of the manuscript. All authors contributed to the article and approved the submitted version.