Coronavirus disease 2019 (COVID-19), also known as the causative novel virus named severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), originated in Wuhan, China, at the end of the year 2019 [1
] and subsequently spread across the globe as a pandemic, affecting 213 countries to date [1
]. As of 16 May 2020, 4,425,485 confirmed cases have been reported, with a mortality rate of 6.8% (302,059 deaths) [1
]. COVID-19 is highly contagious [4
], and there is currently neither a therapeutic agent which showed convincing efficacy, in part due to the inappropriate trial design of most studies [7
], nor a vaccine readily available at present time [8
]. Thus, public health measures have been limited to a focus on curbing the rate of transmission through social distancing.
In the past two decades, the world has already been affected by two severe global human coronavirus (hCoVs) outbreaks: severe acute respiratory syndrome-coronavirus (SARS) [10
] and Middle East respiratory syndrome-coronavirus (MERS) [12
], both exhibiting high mortality rates. These two lethal diseases were caused by SARS-CoV in November 2002 [10
] and MERS-CoV in September 2012 [12
], respectively. It was only after these outbreaks and studying and treating thousands of confirmed cases that we are now in a position to begin to understand the coronavirus [10
]. Rapid deterioration or death from the novel coronavirus is in part attributable to the “cytokine storm” caused by the hCoVs [16
], which is the overproduction of immune cells and cytokines (proteins secreted by the immune cells) that surge into the lungs damaging tissue and organs [19
]. Therefore, treatments have involved drugs that regulate this phenomenon, including corticosteroids, for their known properties of immune regulation and their proven efficacy in treating SARS or MERS [16
However, Russell and colleagues worried about the use of corticosteroids in COVID-19 in a recently published paper [22
]. The authors stated that steroids may be harmful and therefore should not be used. Furthermore, since there have been no randomized control trials (RCTs), or COVID-19 patient data regarding steroid use until recently, the World Health Organization (WHO) and Center for Disease Control (CDC) do not unreservedly recommend the routine use of corticosteroids [23
], which obfuscates clinical practices.
Previous systematic reviews in patients with SARS, MERS or COVID-19 have shown conflicting results. Thus, in this systematic review with the meta-analysis of observational studies, we aim to assess the efficacy of corticosteroids for adult patients admitted to hospital with coronavirus disease.
To our knowledge, this paper is a comprehensive systematic review and meta-analysis providing the most accurate evidence on steroids as an important treatment of choice in critical coronavirus infections. Until now, there were two other systematic reviews [40
] dealing with the correlation between corticosteroid and coronavirus; however, one [40
] reviewed conducted studies without including time or variable-adjusted statistics and the other [41
] only described corticosteroids as an add-on therapy (e.g., combination of ribavirin and corticosteroids). In our study, two reviewers found all the studies from four databases that were missing from the previous study by Russell et al. [22
] to avoid selection biases. There had been no statistical analysis on the efficacy of steroids until the current study.
Of note, several included articles in both categories (“intervention” and “risk factors”) should be interpreted with caution. First of all, steroids were used in more severe cases, and the patients chose to enroll in these included studies. Furthermore, studies did not adjust for confounders related to mortality, such as time, age, or comorbidities. As a result, overall, there were no associations between the use of steroids and mortality (OR 1.152, 95% CI 0.631–2.101, Figure 2
a). However, after adjusting for time or other variables, steroids significantly reduced the risk of mortality (HR 0.378, 95% CI 0.221–0.646 in random effects model, Figure 3
) with similar findings in a fixed-effects model and subset sensitivity analysis. However, when these adjustments were removed after performing a simple chi-square test in both studies, the output was similar to the overall results with no statistically significant differences (p
= 0.158 and p
= 0.304, respectively). Taking these results from two well-executed studies [36
] together, this suggests that well-conducted studies, with time or variable-adjusted statistics, were definitely important in the interpretation of the effect of steroids on lowering mortality. In this regard, prescribing steroids may still be a viable option.
In addition, there were few remaining studies about steroids used in coronavirus after the search process, and most of the studies that ended up being included showed low quality (Tables S6 and S7
and Table 2
). Only two enrolled studies [36
] were ranked as high quality after the search process, and these affected the overall results more than other studies. Excluding low-quality studies changed the final result from no association to better outcome in terms of mortality. We argue that such results from higher quality studies are more appropriate and should be considered relevant before the results of ongoing RCTs become available. As a major point of criticism of the systematic review conducted by Russell et al. [22
], the exclusion of high-quality studies can result in the misinterpretation of the currently available evidence.
Moreover, there have been no systematic reviews or meta-analyses on the timing of initiation of steroids in hCoV. Our study is also the first to collect information on the timing of the initiation of steroids. Since several studies did not mention the timing of steroid use in detail, we examined six studies [33
] and found that steroids were more commonly used in patients who had already reached a critically severe status, such as ALI/ARDS or admission to an ICU. In this case, a large amount of steroids was likely required to suppress the “cytokine storm”.
According to our results, the use of steroids in already critically ill patients in the ICU or on mechanical ventilation can make it difficult to prevent disease progression (OR 1.829, 95% CI 1.018–3.286 in random effects model, Figure 2
c). Moreover, in our “risk factor” study, people who were older and had more comorbidities used a higher amount of steroids along with progression of the disease and showed high mortality (OR 3.133, 95% CI 1.670–5.877 in random effects model, I2
= 0.0%, Figure 4
). Therefore, the use of steroids may be important to prevent disease progression because the “cytokine storm” may not be suppressible when the disease is advanced. It can be assumed that if the patient does not have severe symptoms, the use of steroids in low doses may help to treat coronavirus infection without complications.
Discussions on the use of corticosteroids in coronaviruses such as SARS, MERS, and even COVID-19 have been controversial. Recently, Russell et al. performed a systematic review on several studies about the use of steroids in viral infection and cautioned against the use of steroids [22
]. However, as a result of examining the quality of studies through AMSTAR2 (Table S4
), we found that this paper had potential several critical problems. Among the references cited in this article [22
], a selection bias was evident because complications and side effects of steroids were investigated in studies focusing on steroids only without a retrospective control group. For example, studies on complications such as psychosis [42
], steroid-induced diabetes mellitus [43
], and osteonecrosis [44
] only appear in people who used steroids. The inclusion of studies on influenza [45
] and RSV [46
] can also result in selection bias dealing with different types of viruses. As Ioannidis previously suggested [48
], if there are selection biases and misinterpretation during the selection of studies, the results can misguide treatments and harm patients.
An important shortcoming is found in Russell et al.’s argument [22
] apart from the AMSTAR2 checklist. They did not thoroughly consider the relationship between viral clearance and clinical outcomes. Viral clearance was not significantly associated with clinical outcomes. For SARS, the difference in delayed viral clearance for the treated versus placebo group was two days (17–18 vs. 19–20) [49
], and, for MERS, the viral clearance was not significantly associated with 90-day mortality [35
]. If the delayed clearance of viral RNA is harmful or not is unclear; however, Russell et al. [22
] used it to strengthen their argument.
After Russell et al.’s suggestion, several authors [50
] recommended a short course of corticosteroids at low-to-moderate dose with close monitoring for critically ill patients with COVID-19. Besides, China’s National Health Commission recently developed a modified treatment strategy regarding the use of systematic corticosteroid treatment (methylprednisolone, <1–2 mg per kg body weight, for 3–5 days) for critically ill patients as an adjuvant therapy [53
]. However, the WHO [23
] and the CDC [24
] are yet to change their opinions on the use of steroids based on the study by Russell et al. [22
]. For better global guideline about the use of steroids, additional studies are necessary.
There are several potential limitations of our study that should be mentioned. First, all nine included studies were retrospective cohort studies without any RCTs. Because studies depended on the provided records, reporting bias is possible. Moreover, of the 2140 articles on steroids, only two studies [36
] were adjusted for confounders related to mortality. Until now, there have been no RCTs that have examined the effects of steroids on SARS and MERS based on clinicaltrials.gov, including studies from developing countries. It is surprising that there are no RCTs on this issue for SARS and MERS. As a result, there is a limitation that each result of these included two studies [36
] have be judged the way it is. Second, there can be potential variations of effects for different combinations of steroids (e.g., patients who received both oral prednisolone and intravenous dexamethasone) or other agents such as antiviral agents or antibiotics. Because the included studies were retrospective studies, meta-analyses could not be performed about this point. Third, confounding factors such as rational variations varied and may bias the data, although we made an effort to minimize this bias. Lastly, we could not perform dose-response meta-analyses between steroids and outcomes because of a lack of data.
Nevertheless, we conducted the systematic review and meta-analyses without missing any of the studies previously published. There has been no meta-analysis and evidence-based statistics related to this topic until now. The results from our study suggest that there is a relationship between steroids and better outcomes, especially in the well-controlled studies even though they were not RCTs. This can also be an important key for treatment of COVID-19. In the United Kingdom, the RECOVERY trial recruited 2104 patients in the steroid arm and has recently concluded that dexamethasone in a dosage of 6 mg daily reduced 28-day mortality among patients receiving invasive mechanical ventilation or oxygen at randomization (RR 0.65, 95% CI 0.51–0.82, p
< 0.001; RR 0.80, 95% CI 0.70–0.92; respectively) [54
]. These preliminary findings confirm a report of efficacy of dexamethasone in the management of ARDS [55
]. Other RCTs in COVID-19 have also been registered at clinicaltrials.gov (NCT04244591, NCT04263402, NCT04273321 and NCT04348305) so far and these RCTs are expected to lead to more accurate results on the efficacy of corticosteroids.