Safety and efficacy of glucocorticoids in the treatment of COVID-19: A meta-analysis of randomized control trials

ABSTRACT

Background

It is unclear the efficacy and safety of glucocorticoids compared with placebo or usual care for treatment of COVID-19.

Research design and methods

Randomized controlled trials (RCTs) of corticosteroids in COVID-19 patients from 1 December 2019, to 30 June 2022, were assessed using Cochrane bias risk assessment method and improved Jadad score scale. GRADEpro was used to rate the quality of evidence for outcomes.

Results

Fifteen RCTs were included, including 10,620 patients. Glucocorticoid treatment for severe and critical COVID-19 showed lesser all-cause mortality (OR = 0.85, 95% CI [0.76, 0.94], P = 0.002) than conventional treatment. However, for mildly ill patients, neither inhaled drugs nor intravenous drugs reduced mortality (OR = 0.64, 95% CI [0.24, 1.76], P = 0.39). Glucocorticoids had no significant effect on the adverse reactions of patients (OR = 1.18, 95% CI [0.77, 1.80], P = 0.44) compared with usual care/placebo. Subgroup analysis demonstrated that dexamethasone significantly reduced the mortality of COVID-19 patients. Low-dose glucocorticoids were also associated with lower all-cause mortality.

Conclusion

Glucocorticoids (especially dexamethasone) reduce mortality of patients with severe and critical COVID-19 with no significant effect on the incidence of adverse reactions (moderate quality). In contrast, glucocorticoids do not benefit patients with mild symptoms (low quality).

KEYWORDS:

• Glucocorticoid
• dexamethasone
• COVID-19
• adverse reaction
• severe type
• meta-analysis

1. Introduction

Coronavirus disease 2019 (COVID-19) caused by a new coronavirus named severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) [1] has triggered a global health crisis [2]. The risk of infection is high, the strain mutates quickly, and the prognosis is poor for severely ill patients. People aged ≥ 60 years and those with underlying conditions such as hypertension, heart and lung problems, diabetes, obesity, or cancer have a higher risk of developing severe COVID-19. However, individuals of any age can develop the disease, become seriously ill, or die from COVID-19 [1]. In the latest update in September 2022, more than 600 million COVID-19 infections and more than 6 million COVID-19-related deaths have been recorded worldwide [3].

SARS-CoV-2 enters host cells by binding to angiotensin-converting enzyme 2 (ACE2) after cleavage by transmembrane protease serine 2 (TMPRSS2) [4]. Its pathogenesis is closely related to an immune-mediated cytokine storm [5]. Glucocorticoids, as the standard therapy for inflammatory and autoimmune disorders, have a good inhibitory effect on inflammatory factors and are often used as adjuvant therapy for viral pneumonia [6]. Its main anti-inflammatory effect is to inhibit the inflammatory process and restore balance in vivo by the inhibition of a large number of pro-inflammatory genes encoding cytokines, chemokines, cell adhesion molecules, inflammatory enzymes, and receptors [7]. Hence, glucocorticoids can play a central role in immunosuppression during an acute cytokine storm for the symptomatic treatment of COVID-19.

Studies on the efficacy and safety of glucocorticoids in the treatment of COVID-19 emerge endlessly. WHO guidelines [8] strongly recommend systemic glucocorticoid treatment for severe and critical COVID-19 patients. Wagner C et al. [9] reached a similar conclusion that systemic glucocorticoid treatment is beneficial to reduce all-cause mortality in severe and critical diseases. With the development of time and the variation of the virus, more and more patients with mild symptoms are diagnosed. However, there is a lack of efficacy evaluation for asymptomatic patients or patients with mild symptoms, and the therapeutic effect of inhaled glucocorticoid on COVID-19 is still unclear. Therefore, it is necessary to conduct a systematic analysis again. This study summarizes RCTs carried out globally with both systemic and inhaled glucocorticoids. In addition to basic indicators such as mortality and adverse reactions, subgroup analysis was also conducted in terms of disease degree, drug type and drug dose. The goal of this study is to comprehensively assess the efficacy and safety of glucocorticoids versus usual care or placebo for COVID-19.

2. Methods

2.1. Registration and protocol

This meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement [10] and registered with PROSPERO (CRD42020193268). The PRISMA checklist and the review protocol can be accessed in Appendix 1.

2.2. Search strategy

We performed a systematic literature search in electronic databases, including PubMed, Embase, the American clinical trial database, and the Chinese database CNKI for eligible studies published until 30 June 2022. There was no language restriction. The search strategy was based on the following keywords and MeSH terms: ‘glucocorticoid, prednisolone, dexamethasone, budesonide, hydrocortisone, methylprednisolone, prednisone, COVID-19, SARS-CoV-2, and randomized controlled trials’. Detailed search strategy can be found in Appendix 2. Two investigators (Ye Zhang, Dian Li) independently performed literature screening to identify eligible studies. Disagreements were resolved by consensus. The reference lists of the relevant articles were also reviewed to identify potential studies.

2.3. Protocol and guidance

The included studies met the following criteria: (1) The included patients were clinically suspected or laboratory diagnosed as having COVID-19. (2) They were randomized controlled trials. (3) Intervention: The experimental group was administered glucocorticoids of any type in any form, dose, or duration based on standard treatment, while the control group was given standard treatment or placebo. (4) After the completion of the experiment, accurate data were obtained.

2.4. Exclusion criteria

(1) Non-randomized controlled trials (cohort studies, case-control studies, etc.) (2) Trials with unavailable data and trials with missing data without explanation or treatment. (3) Low-quality studies (including less-than-rigorous design and small enrollment) (4) Studies that repeatedly reported the same cohort of patients.

2.5. Assessment of risk of bias

The Cochrane Risk of Bias Assessment Method and the Modified Jadad Rating Scale were used to assess the risk of bias for the included studies. Funnel chart were used to evaluate publication bias. Disagreements were settled through negotiation or according to the opinions of a third party.

2.6. Outcomes

The main outcome was all-cause mortality, and subgroup analyses were conducted according to the type of glucocorticoid drug administered, doses of glucocorticoids and the severity of the patient’s condition. According to Guidelines for clinical application of glucocorticoids [11], we defined > 60 mg prednisone equivalent a day as high dose of systemic glucocorticoid, and < 60 mg/d as low dose; According to the Guidelines for the Prevention and Treatment of Bronchial Asthma (2020 Edition) [12], >800ug/d is defined as high dose of budesonide, and < 800ug/d is low dose; > 320ug/d is defined as the high dose of ciclesonide, and < 320ug/d is low dose. The severity of COVID-19 is determined according to the WHO guidelines [13]. We analyzed the included studies. Patients with mild and moderate disease exhibited only symptoms related to COVID-19 (low fever, mild fatigue, dysosmia, and/or pneumonia) and severe and critical patients were characterized by dyspnea and/or hypoxemia, which are consistent with the definition of the guidelines.

Secondary outcomes included adverse effects (defined as the number of patients with any event), length of hospital stay, and the time to conversion/conversion rate of pharyngeal swab nucleic acid testing.

2.7. Assessment of the quality of evidence

We followed the GRADE guidelines [14] and used GRADEpro 3.6 to rate the quality of evidence for the outcomes included in this study. The evaluation of evidence grade is divided into high (no degradation), moderate (reduced by 1 grade), low (reduced by 2 grades) and very low (reduced by 3 grades) based on considerations of five factors – risk of bias, consistency, indirectness, immediacy, and publication bias. Wenxiao Qiao and Dian Li rated each outcome and resolved discrepancies by consensus.

2.8. Data synthesis

Using the software ReviewManage 5.4.1, statistical analysis of the gathered data was performed. The odds ratio (OR) was used as the effect size for the count data and the confidence interval (CI) was set at 95%. The heterogeneity of the included studies was examined using x² test, and if I² < 50% and Paired (P) > 0.1, the heterogeneity of the included studies was considered not statistically significant, and the fixed-effects model could be used to combine the effect sizes. If I² ≥ 50% and/or P < 0.1, this suggested actual heterogeneity among the included studies. A random-effects model was chosen to combine the effect sizes, or a sensitivity analysis was chosen for the analysis and investigation of the causes of heterogeneity and to exclude literature that caused greater heterogeneity. If a meta-analysis of data from the literature was not possible, descriptive analysis was used instead.

3. Results

3.1. Eligible studies and study characteristics

We initially identified 488 studies and included 15 eligible trials [15–29] in the final meta-analysis. The trials comprised 10,620 participants, including 4090 patients in the glucocorticoid group and 6530 patients in the conventional treatment group. The literature screening process and results are demonstrated in Figure 1. The details of the included trials are summarized in Table 1, including the number of participants, sex, average age, drug use, dose, main outcome of the experimental group, and control group.

Figure 1. Study selection process.

Table 1. Included trails and characteristics.

Author	Published time	Country	Design	Age	Mean age ((male/female)	Sex(male/female)	No. of participants(experiment/control)	Control	Glucocorticoid species	Dose and duriation	Inclusion criteria	The main outcome
Ramakrishnan et al [16].	2021	Britain	open-label, parallel-group, phase 2, randomized controlled trial	>18	44/46	59/80	139 (70/69)	Usual care	Inhaled budesonide	400 μg twice daily until symptom resolution	Adults who have developed COVID-19 symptoms within 7 days (new-onset cough and fever or loss of smell, or both).	COVID-19-related urgent care visits, including emergency department assessment or hospitalization.
Yu et al [17].	2021	Britain	multicentre, open-label, multi-arm, randomized, controlled, adaptive platform trial	>50	64.7/63.8	Unclear	1856 (787/1069)	Usual care	Inhaled budesonide	800 μg twice daily for 14 days	Eligible participants felt unwell for up to 14 days with suspected COVID-19 but not admitted to hospital.	Time to first self-reported recovery and hospital admission or death related to COVID-19, within 28 days
Ezer et al [18].	2021	Canada	randomized, double blind, placebo controlled trial	>18	32/35	Unclear	203 (105/98)	Placebo	Inhaled and intranasal ciclesonide	inhaled ciclesonide (600 μg twice daily) and intranasal ciclesonide (200 μg daily)	PCR confirmed COVID-19, presenting with fever, cough, or dyspnea.	Symptom resolution at day 7
Clemency et al [19].	2022	America	phase 3, multicenter, double-blind, randomized clinical trial	>12	43.7/42.9	179/221	400 (197/203)	Placebo	Inhaled Ciclesonide	160 μg twice daily	(1)PCR confirmed COVID-19 during the previous 72 hours, presenting with fever, cough, or dyspnea. (2) were not hospitalized or under consideration for hospitalization, (3) Spo2 < 93%.	Time to alleviation of all COVID-19–related symptoms (cough, dyspnea, chills, feeling feverish, repeated shaking with chills, muscle pain, headache, sore throat, and new loss of taste or smell) by day 30.
Jamaati et al [20].	2021	Iran	randomized, controlled, clinical trial	>18	62/62	36/24	50 (25/25)	Usual care	Intravenous dexamethasone	20 mg/day from day 1–5 and then 10 mg/day from day 6–10.	(1) PCR confirmed COVID-19; (2) PaO2/FiO2 were between 100 and 300 mmHg; (3) bilateral lung infiltration.	The need for invasive mechanical ventilation and death rate
Tomazini et al [21].	2020	Baxi	multicenter, randomized, open-label, clinical trial	>18	60.1/62.7	187/112	299 (151/148)	Standard care	Intravenous dexamethasone	20 mg daily for 5 days, 10 mg daily for 5 days or until ICU discharge,	PCR confirmed COVID-19, and received mechanical ventilation within 48 hours of meeting criteria for moderate to severe ARDS with PaO2:FIO2 < 200 mmHg.	Ventilator-free days during the first 28 days, defined as being alive and free from mechanical ventilation.
Ghanei et al [22].	2021	America	multicenter, randomized open-label trial	>16	58.2/57.6	112/114	226 (116/110)	Azithromycin	Prednisolone	25 mg daily for 5 days	PCR confirmed COVID-19 and Spo2 < 94%	Admission to ICU
Tang et al [23].	2021	Beijing	prospective, multicenter, single-blind, randomized control trial	>18	57/55	41/45	86 (43/43)	Usual care	Intravenous methylprednisolone	1 mg/kg daily for 7 days	Confirmed COVID-19.	The incidence of clinical deterioration 14 days after randomization.
Edalatifard et al. [24]	2020	Iran	single-blind, randomized, controlled, clinical trial	>18	55.8/61.7	39/23	62 (34/28)	Standard care	Intravenous methylprednisolone	250 mg/day for 3 days	Confirmed COVID-19 with Spo2 < 90%, CRP >10, and interleukin-6 (>6) at the early pulmonary phase of disease before connecting to the ventilator and intubation.	The time of clinical improvement or death, whichever came first.
Jeronimo et al [25].	2021	Baxi	randomized, Double-Blind, Phase IIb, Placebo-Controlled Trial	≥ 18	54/57	254/139	393 (194/199)	Placebo	Intravenous methylprednisolone	0.5 mg/kg twice daily for 5 days	Hospitalized patients with clinical, epidemiological and/or radiological suspected COVID-19	28-day mortality
Luis Corral-Gudino et al [26].	2021	Spain	randomized, open-label, controlled, two-arm, parallel-group, trial	>18	73/66	39/25	64 (35/29)	Standard care	Intravenous methylprednisolone	40 mg bid for 3 days followed by 20 mg bid for 3 days	PCR confirmed COVID-19.	Composite endpoint, including in-hospital all-cause mortality, escalation to ICU admission, or progression of respiratory insufficiency that required NIV.
Dequin et al [27].	2020	France	multicenter randomized double-blind sequential trial	≥ 18	54/50	105/45	149 (76/73)	Placebo	low-dose hydrocortisone	200 mg/d until day 7 and then decreased to 100 mg/d for 4 days and 50 mg/d for 3 days, for a total of 14 days	PCR confirmed COVID-19.	Treatment failure on day 21, defined as death or persistent dependency on mechanical ventilation or high-flow oxygen therapy.
Angus et al [28].	2020	multicenter	randomized controlled, ongoing, international, multicenter, open label trial	≥ 18	60.4/59.9	170/68	238 (137/101)	Usual care	Intravenous hydrocortisone	50 mg, every 6 hours for 7 days	Patients with a presumed or confirmed SARS-CoV-2 infection are admitted to ICU to provide respiratory or cardiovascular organ support.	Organ support–free days (days alive and free of ICU-based respiratory or cardiovascular support) within 21 days.
Marie et al [29].	2021	Denmark	multicentre, parallel-group, placebo-controlled, blinded, centrally randomized, stratified clinical trial	≥ 18	59/62	24/6	30 (16/14)	Placebo	Intravenous hydrocortisone	200 mg/d for 7 days	Adults with a confirmed diagnosis of COVID-19 and severe hypoxia (using mechanical ventilation or supplemental oxygen, with a flow rate of at least 10 L/min)	The number of days alive without life support at day 28

PaO2/FiO2: ratio of partial pressure of oxygen in arterial blood to fraction of inspired oxygen; ICU: the intensive care unit; Spo2: oxygen saturation; CRP: C-reactive protein; NIV: noninvasive ventilation

3.2. Assessment of risk of bias

The Cochrane risk-of-bias assessment was conducted from six perspectives: random, allocation, blinding, measurement, reporting, and other sources of bias. The results demonstrated that the overall quality of the included studies was high and the risk of bias was low. The bias was mainly focused on the implementation of the blind method as some RCTs chose open research due to ethical reasons (Figure 2). The Jadad modified rating scale evaluation demonstrated that 13 articles were of high quality and two articles were of low quality (Table 2).

Figure 2. Risk of bias summary.

Table 2. The Jadad modified rating scale.

RCT	Random	Allocation	Blinding	Loss of follow-up	Score
Ramakrishnan	2	0	0	1	3
Yu	2	2	0	1	5
Ezer	2	1	1	1	5
Clemency	2	2	1	1	6
Jamaati	1	0	0	1	2
RECOVERY	2	2	0	1	5
Tomazini	2	2	0	1	5
Ghanei	2	2	0	1	5
Tang	2	2	1	1	6
Edalatifard,	2	1	1	1	5
Jeronimo	2	0	1	1	4
Luis Corral-Gudino	2	2	0	1	5
Dequin	2	2	1	1	6
Angus	1	1	0	1	3
Marie	2	2	1	1	6

The score is 1–3 for low quality literature and 4–7 for high quality literature.

3.3. Primary outcome: all-cause mortality

Fourteen trials [15,17–29] reported all-cause mortality. Among the 4020 patients who received glucocorticoid treatment, 732 died (18.2%), while 1382 of 6461 patients in the control group died (21.4%). There was a statistically significant difference between the glucocorticoid group and the control group (OR = 0.85, 95% CI [0.76, 0.94], I² = 24%, P = 0.002; Figure 3), indicating that glucocorticoids can effectively reduce the mortality of COVID-19 patients, compared with usual care or placebo treatment. The sensitivity analysis was robust, and the conclusion was stable.

Figure 3. All-cause mortality.

3.4. Subgroup analyses

3.4.1. Type of glucocorticoids

Six types of glucocorticoids were included in this study: dexamethasone, methylprednisolone, prednisolone, hydrocortisone, budesonide, and ciclesonide; to investigate the efficacy of these different types of glucocorticoids in adjuvant therapy of novel coronavirus, the following analyses were performed.

Three trials used dexamethasone [15,20,21], with a total of 6774 patients, and the heterogeneity among the included studies was low (I² = 0%, P = 0.83). The fixed effects model was selected to combine the effect size for meta-analysis, and the results demonstrated that OR = 0.86, 95% CI (0.76–0.97), P = 0.01 (Supplemental Figure 1A). The difference was statistically significant, indicating that dexamethasone adjuvant therapy can significantly reduce the mortality of patients with COVID-19.

No mortality was reported in two RCTs using ciclesonide [18,19]. Hence, the disease remission rate (the resolution of all COVID-19-related symptoms by day 30) was used for evaluation. Heterogeneity test yielded a result of I² = 0, indicating that the number is completely homogeneous. The use of a fixed effect model combined with the effect of meta-analysis suggested that the effect of ciclesonide adjuvant therapy on the improvement of symptoms was not statistically significant (OR = 1.40, 95% CI [1.01, 1.96], P = 0.05, Supplemental Figure 1B).

Two trials used budesonide [16,17], but only one reported all-cause mortality; therefore, the disease deterioration rate (COVID-19-related urgent care visits, including emergency department assessment or hospitalization) was selected for evaluation. Heterogeneity between the studies was high (I² = 64%). When the random effects model was used to analyze the effect, we obtained OR = 0.38, 95% CI (0.08, 1.75), and P = 0.21 (Supplemental Figure 1C), suggesting that budesonide adjuvant therapy had no significant effect on the deterioration of COVID-19 patients compared with conventional therapy or placebo.

In addition, one RCT [22] used prednisolone as an adjuvant therapy, with the result P = 0.47. Four articles [23–26] used methylprednisolone], and meta-analyze showed OR = 0.54, 95% CI (0.18, 1.63), and P = 0.27. Three trails [27–29] used hydrocortisone [and meta-analyze showed OR = 0.80, 95% CI (0.37, 1.72), and P = 0.57. Subgroup analysis demonstrated that there was no significant difference in mortality between the three drugs and the control group (Supplemental Figure 1A)

3.4.2. Different doses of glucocorticoids

Nine RCTs [15,19,22,23,25–29] used low-dose dexamethasone, the heterogeneity test I² = 0, and the fixed effect model combined with the effect quantity was used for meta-analysis, suggesting that compared with usual care or placebo, low dose glucocorticoid is beneficial to reduce mortality (OR = 0.86, 95% CI [0.77, 0.96], P = 0.009, Supplemental Figure 2A). Five RCTs [17,18,20,21,24] used high-dose glucocorticoid, and the heterogeneity between studies was high (I² = 62%). Using a random effect model combined with the effect amount, the OR = 0.61, 95% CI (0.28, 1.34), P = 0.22 (Supplemental Figure 2B) were obtained, suggesting that high-dose glucocorticoid has no significant impact on the mortality of patients with COVID-19.

3.4.3. Different disease processes

This study analyzed the effect of glucocorticoid adjuvant therapy from the perspectives of mild, moderate, severe, and critical types of disease.

The RECOVERY study included more comprehensive cases, which were divided into mild and moderate type (1535 cases) and severe type (4890 cases). A total of 6386 severe and critically ill patients from seven RCTs [15,20–22,25,27,28] were included in this study. There was no statistical difference in the heterogeneity test (I² = 0). The meta- analysis demonstrated that OR = 0.78, 95% CI [0.69, 0.88], P < 0.0001 (Supplemental Figure 3). These results suggest that glucocorticoid adjuvant therapy has statistical significance in reducing mortality in patients with severe and critical COVID-19.

A total of 3616 patients with mild and moderate disease from four RCTs [15–17,23] were included in the mortality analysis; the glucocorticoid mortality rate was 6.9% (96/1401) and the control group mortality rate was 7.5% (167/2215), which was not statistically significant (OR = 0.64, 95% CI (0.24, 1.76), P = 0.39; Supplemental Figure 3).

3.5. Secondary outcome

3.5.1. Adverse reactions

A total of 11 RCT’s [17–19,21,23–29] were included in this study to record the adverse reactions during adjuvant therapy with glucocorticoids. Specific types of adverse reactions can be found in Table 3. Adverse effects occurred in 269 of the 1895 patients treated with glucocorticoids (14.2%) and 240 of the 1890 patients in the control group (12.7%). The random effects model combined with effect size was used for analysis, and the results demonstrated that OR = 1.88, 95% CI (0.77, 1.80), P = 0.44 (Figure 4a), indicating that there was no statistically significant difference in adverse reactions between glucocorticoid adjuvant therapy and conventional treatment.

Figure 4. Adverse reactions.

Table 3. Types of adverse reactions.

	Illustrative comparative risks* (95% CI)
Outcomes	Assumed riskControl			Corresponding risk Glucocorticoid			Relative effect(95% CI)			No of Participants(studies)				Quality of the evidence(GRADE)	Comments
All-cause mortality
Patient or population: patients with COVID-19 Settings: all-cause mortality Intervention: glucocorticoid
All-cause mortality	Study population						OR 0.85 (0.76 to 0.94)			10,481(14 studies)				⊕⊕⊕⊕high
	214 per 1000			188 per 1000(171 to 204)
	Moderate
	215 per 1000			189 per 1000(172 to 205)
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).CI: Confidence interval; OR: Odds ratio;
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.
Adverse reaction
Patient or population: patients with COVID-19 Settings: Intervention: glucocorticoid
Adverse reactions		Study population						OR 1.18 (0.77 to 1.8)			3875(11 studies)		⊕⊕⊕⊝moderate^1
		121 per 1000			140 per 1000(96 to 199)
		Moderate
		71 per 1000			83 per 1000(56 to 121)
Hyperglycemia		Study population						OR 1.25 (0.67 to 2.33)			796(4 studies)		⊕⊕⊝⊝low^2
		343 per 1000			395 per 1000(259 to 548)
		Moderate
		223 per 1000			264 per 1000(161 to 401)
Infection		Study population						OR 0.85 (0.53 to 1.37)			511(4 studies)		⊕⊕⊕⊝moderate^2
		181 per 1000			159 per 1000(105 to 233)
		Moderate
		29 per 1000			25 per 1000(16 to 39)
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).CI: Confidence interval; OR: Odds ratio;
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.
^1 The funnel plot was asymmetrical. ^2 The number of included studies is small.
Type of glucocorticoids
Patient or population: patients with COVID-19 Settings: Intervention: glucocorticoids
Dexamethasone			Study population						OR 0.86 (0.76 to 0.97)			6774(3 studies)		⊕⊕⊕⊝moderate^1
			271 per 1000			242 per 1000(220 to 265)
			Moderate
			600 per 1000			563 per 1000(533 to 593)
Prednisolone			Study population						OR 0.62 (0.17 to 2.26)			226(1 study)		⊕⊝⊝⊝very low^1,2,3
			55 per 1000			35 per 1000(10 to 115)
			Moderate
			55 per 1000			35 per 1000(10 to 116)
Methylprednisolone			Study population						OR 0.54 (0.18 to 1.63)			603(4 studies)		⊕⊕⊝⊝low^2
			314 per 1000			198 per 1000(76 to 428)
			Moderate
			277 per 1000			171 per 1000(65 to 384)
Hydrocortisone			Study population						OR 0.8 (0.37 to 1.72)			558(3 studies)		⊕⊕⊝⊝low^1,4,5
			293 per 1000			249 per 1000(133 to 416)
			Moderate
			274 per 1000			232 per 1000(123 to 394)
Budesonide			Study population						OR 0.38 (0.08 to 1.75)			1995(2 studies)		⊕⊕⊝⊝low^1,4,5
			19 per 1000			7 per 1000(2 to 33)
			Moderate
			85 per 1000			34 per 1000(7 to 140)
Ciclesonide			Study population						OR 1.4 (1.01 to 1.96)			603(2 studies)		⊕⊕⊝⊝low^1,4
			575 per 1000			654 per 1000(577 to 726)
			Moderate
			542 per 1000			624 per 1000(544 to 699)
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).CI: Confidence interval; OR: Odds ratio;
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.
^1 The number of included studies is small.^2 The heterogeneity is high(>50%).^3 Wide confidence interval.^4 Blinding missing and no allocation concealment.^5 The heterogeneity is high, which may be caused by the lack of blinding, so there is no repeated degradation.
Different disease processes
Patient or population: patients with COVID-19 Settings: Intervention: glucocorticoid
Mild and ordinary			Study population						OR 0.64 (0.24 to 1.76)			3616(4 studies)		⊕⊕⊝⊝low^1,2
			75 per 1000			50 per 1000(19 to 126)
			Moderate
			82 per 1000			54 per 1000(21 to 136)
Severe and critically			Study population						OR 0.78 (0.69 to 0.88)			6386(7 studies)		⊕⊕⊕⊝moderate^2
			306 per 1000			256 per 1000(233 to 279)
			Moderate
			327 per 1000			275 per 1000(251 to 300)
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).CI: Confidence interval; OR: Odds ratio;
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.
^1 The heterogeneity is high, which may be caused by the lack of blinding, so there is no repeated degradation.^2 The number of included studies is small.
Different doses of glucocorticoids
Patient or population: patients with COVID-19 Settings: Intervention: Different doses of glucocorticoids
Low-dose			Study population						OR 0.86 (0.77 to 0.96)			8011(9 studies)		⊕⊕⊕⊝moderate^1
			246 per 1000			219 per 1000(201 to 239)
			Moderate
			172 per 1000			152 per 1000(138 to 166)
High-dose			Study population						OR 0.61 (0.28 to 1.34)			2470(5 studies)		⊕⊕⊝⊝low^2,3,4
			94 per 1000			60 per 1000(28 to 122)
			Moderate
			429 per 1000			314 per 1000(174 to 502)
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).CI: Confidence interval; OR: Odds ratio;
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.
^1 The funnel plot was asymmetrical. ^2 Blinding missing and no allocation concealment.^3 The heterogeneity is high, which may be caused by the lack of blinding, so there is no repeated degradation.^4 The number of included studies is small.

Among the adverse reactions mentioned, hyperglycemia and infection are the most common. Four RCTs [21,23,25,26] mentioned hyperglycemia events, and we carried out a meta-analysis. The results showed that the heterogeneity was high (I² = 58%). The random effect model combined with the analysis of the effect quantity obtained OR = 1.25, 95% CI (0.67, 2.33), P = 0.49 (Figure 4b), suggesting that glucocorticoids had no significant effect on the occurrence of hyperglycemia symptoms. Four trails [21,23,24,26] mentioned infection and the meta-analysis showed that the heterogeneity was low (I² = 23%, P = 0.27). The fixed effect model combined with the effect quantity was analyzed, and the OR = 0.85, 95% CI (0.53, 1.37), P = 0.50 (Figure 4c).

3.5.2. The length of hospital stay

Three RCT’s described the effect of adjuvant glucocorticoid therapy on the length of hospital stay (Table 4). Jamaati et al. [20] compared the median length of hospital stay between patients who received dexamethasone adjuvant therapy and those who received conventional therapy (11 days in the dexamethasone group versus 6 days in the conventional treatment group, P = 0.036). The length of hospital stay was prolonged for patients who received dexamethasone adjuvant therapy. Ghanei et al. [22] investigated the effect of prednisolone on the average length of hospital stay and conducted modified intention-to-treat (ITT) and per-protocol (PP) analyses. There was no significant difference between prednisolone use and ITT analysis (5.5 days in the experimental group and 6.4 days in the control group, P = 0.09). PP analysis (experimental group 4.4 days, control group 5.9 days, P = 0.03) was statistically significant. Jeronimo et al. [25] included 394 cases, and the median length of hospital stay was 9 days in the conventional treatment group and 10 days in the methylprednisolone group, P = 0.296; the difference was not statistically significant.

Table 4. The length of hospital stay in patients with COVID-19.

RCT	Usual care		Glucocorticoid		P
	Median	Average	Median	Average
Jamaati	6 (4–9)		11 (6–16)		0.036
Ghanei		ITT:6.4 PP:5.9		ITT:5.5 PP:4.4	0.09 0.03
Jeronimo	9 (7–12)		10 (9–13)		0.296

ITT: Intention-to-treat analysis, PP: Per-protocol analysis

3.5.3. The time of nucleic acid test turning negative

Tang et al. [23] included 86 patients with common COVID-19 and discovered that the duration of throat viral RNA detection in the methylprednisolone group was 11 days (interquartile range, 6–16 days), which was significantly longer than that in the control group (8 days, 2–12 days), P = 0.03; this was statistically significant. Hence, glucocorticoid adjuvant therapy can prolong the time of the nucleic acid test turning negative. Jeronimo et al. [25] researched the negative conversion rate of nucleic acid detection on the 5th and 14th days. On the 5th day, the negative conversion rate of throat swabs in the conventional treatment group was 73/139 (52.5%), while that in the methylprednisolone group was 75/144 (52.1%), P = 0.942; the difference was not statistically significant. On the 7th day, the rate of throat swab turning negative was 45/95 (47.3%) in the conventional treatment group and 56/117 (47.9%) in the methylprednisolone group, P = 0.943, which was not statistically significant, indicating that methylprednisolone adjuvant therapy had no significant effect on the negative conversion of nucleic acid detection.

3.6. Publication bias

Only two outcomes, all-cause mortality and adverse reactions, were included in more than 10 articles. The funnel plot was selected to evaluate the risk of publication bias. The funnel plot of mortality (Figure 5a) is almost symmetrical, indicating that there is no risk of publication bias. The funnel plot of adverse reactions (Figure 5b) is asymmetric, indicating the high risk of publication bias.

Figure 5. Funnel plots.

3.7. GRADE evidence quality evaluation results

GRADE results showed that the all-cause mortality was high quality and the adverse reactions were moderate quality. Among the 12 outcomes of subgroup analysis, 4 results were moderate quality, 7 results were low quality, and 1 result was very low quality. The results are shown in Table 5.

Table 5. Types of adverse reactions.

		Types of adverse reactions
Author	Glucocorticoid species	Glucocorticoid	Control
Ramakrishnan et al. [16]	Inhaled budesonide	Sore throat; dizziness	Specific name not reported
Yu et al. [17]	Inhaled budesonide	Hospital admission unrelated to COVID-19	Hospital admission unrelated to COVID-19
Ezer et al. [18]	Inhaled Ciclesonide	Headache; nausea	Headache; nausea
Clemency et al. [19]	Inhaled Ciclesonide	Headache; Dry mouth	Headache, dry mouth
Tomazini et al. [21]	Intravenous dexamethasone	Hyperglycemia; secondary infection	Hyperglycemia; secondary infection
Tang et al. [23]	Intravenous methylprednisolone	Hyperglycemia; pneumonia	Hyperglycemia; pneumonia
Edalatifard et al. [24]	Intravenous methylprednisolone	Infection; edema	Shock
Jeronimo et al. [25]	Intravenous methylprednisolone	Hyperglycemia	Hyperglycemia
Luis Corral-Gudino et al. [26]	Intravenous methylprednisolone	Hyperglycemia; Nosocomial infection	Hyperglycemia, Nosocomial infection
Dequin et al. [27]	Intravenous hydrocortisone	Intra-abdominal hemorrhage; cardiac arrest; cerebral vasculitis	No adverse reaction
Angus et al. [28]	Intravenous hydrocortisone	Severe neuromyopathy; fungemia	Specific name not reported

4. Discussion

Previous studies differed in conclusions regarding all-cause mortality [30,31], which probably due to the inclusion of nonrandomized controlled trials. WHO Rapid Evidence Appraisal for COVID-19 Therapies (REACT) Working Group [32] analyzed RCTs, and the research results showed that systemic glucocorticoid was associated with lower 28-day mortality. Siemieniuk et al. [33] also suggested that compared with usual care or placebo, glucocorticoids may reduce the mortality of severe patients, and our research results were consistent with them. Subgroup analysis showed that it was not beneficial for ordinary and mild patients. In short, the use of glucocorticoids was more effective in patients with severe COVID-19 who had more severe inflammation and cytokine storm [34].

Based on subgroup analysis, the efficacy of glucocorticoids was related to the type of drug. Only dexamethasone administration had a significant effect on the reduction of patient mortality, consistent with a previous study [35]. Even though Edalatifard at al [24]. reported that high-dose methylprednisolone was beneficial in reducing mortality, there was no data from other RCTs to verify it. The dose of glucocorticoid affects the curative effect. The results of meta-analysis showed that low dose of glucocorticoid may be related to the reduction of all-cause mortality, but high dose of glucocorticoid has nothing to do with it. The efficacy of glucocorticoids may also be related to the dosage form and the severity of the disease. Both ciclesonide and budesonide are inhaled drugs, and the main purpose of treatment is to relieve the patient’s condition, such as cough, asthma, loss of sense of smell, etc. The results of this study show that the above two drugs are not beneficial to the relief of symptoms. Hence, it is not recommended to use inhaled drugs to relieve the clinical symptoms of mild and common patients. For severe patients, no RCT uses inhaled glucocorticoids, so it is regrettable that its efficacy for severe patients cannot be evaluated, and more RCT results are needed to verify.

The literatures included in this study used short-term use of glucocorticoids (less than 15 days). Studies have shown that even short-term use can cause a variety of side effects [36], such as hyperglycemia, hypertension, arrhythmia, and decreased immune function. This is also a concern for patients. Our study analyzed 3874 patients in 11 RCTs, and found that compared with conventional treatment, the probability of adverse reactions in the glucocorticoid group was not statistically different. When patients have been using glucocorticoids to treat underlying diseases (such as regular use of budesonide for asthma), they can continue to use them after being infected with COVID-19 without causing a greater burden on the body. However, even without COVID-19, the adverse reactions such as hypertension, hyperglycemia, and gastrointestinal reactions are common after long-term use of glucocorticoids [37]. Hence, if a patient suffers from these preexisting conditions, regular monitoring and timely adjustment of medication is needed.

In addition, we found that patients who used inhaled glucocorticoids generally had mild adverse reactions such as headache, which may be related to their mild illness. However, severe patients who used systemic glucocorticoids (such as dexamethasone and methylprednisolone) were more prone to hyperglycemia and infection. Meta-analysis results showed that there was no statistical difference in hyperglycemia and infection between glucocorticoid group and usual care or placebo group. It is important to note that this study lacks long-term follow-up data of patients, and it is impossible to know the long-term adverse reactions, so further research is needed.

According to the results of the three included RCTs, glucocorticoids did not shorten the time for negative nucleic acid conversion in throat swabs, which may be related to the suppression of immune responses with the early use of glucocorticoids [23].

This study had some limitations. First, there was no uniform standard for the time and dose of hormones used in various studies. Because of the limited RCTs and heterogeneity between trials, it was not possible to perform meta-analysis on other outcome indexes, such as length of hospitalization, nucleic acid detection, negative rate, negative duration, computed tomography (CT) findings, etc. Because of the release of RECOVERY’s study data [15], the use of glucocorticoids in critically ill patients is highly recommended by the WHO, and many trials have been forced to stop recruiting patients without collecting the expected sample size. Due to the rapid mutation of SARS-COV-2, some studies have not been published, which will lead to publication bias. Finally, in search process, in order to avoid duplicate documents, we did not include Cochrane database in the search scope, which may lead to data loss.

In addition, as the SARS-CoV-2 virus continues to mutate over time, the likely role of glucocorticoids on different virus mutations can be used as further research in the future.

5. Conclusions

In summary, glucocorticoids, especially dexamethasone, can reduce the mortality of critically ill patients with severe COVID-19 (moderate quality). Low-dose glucocorticoids are associated with lower all-cause mortality (moderate quality). No matter inhaled or systemic glucocorticoid, it has no significant effect on the disease remission rate and mortality in patients with mild and common COVID-19 (low quality). Compared with the control group, there was no significant difference in adverse reactions caused by glucocorticoids, indicating that there was no obvious increase in adverse reactions to some extent (moderate quality). Other outcomes were low quality and shall be carefully explained. Limited by the number and quality of the included studies, hospitalization time, nucleic acid detection, and negative time could not be meta-analyzed, and more high-quality RCTs need to be carried out to verify this.

Declaration of interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/17476348.2023.2177155

Additional information

Funding

This work was supported by the Shanxi Province Science and Technology Research (202102130501004).

Appendix 1

Search strategies

Pubmed

Search: ((((‘COVID-19’[Mesh]) AND ‘Glucocorticoids’[Mesh] AND (‘2019/12/01’[Date – Publication]: ‘2022/06/30’[Date – Publication])) NOT (meta analysis[Title])) NOT (review[Title])) NOT (systematic review[Title])

Embase:

(‘glucocorticoids’/exp OR glucocorticoids) AND (‘covid 19’/exp OR ‘covid 19’) AND ‘randomized controlled trial (topic)’ AND [01-12-2019]/sd NOT [01–07-2022]/sd AND [2019–2022]/py NOT ‘systematic review’ NOT ‘review’ NOT ‘meta analysis’

CNKI:

(Subject = (glucocorticoid + prednisolone + dexamethasone + budesonide + hydrocortisone + methylprednisolone + prednisolone)) AND (Subject = (novel coronavirus pneumonia + COVID-19)) AND (Title, Keyword and Abstract = randomization + trial)

ClinicalTrials.gov

Glucocorticoids AND Completed Studies AND Studies With Results AND Interventional Studies AND COVID-19