Intravenous methylprednisolone pulses in hospitalised patients with severe COVID-19 pneumonia: a double-blind, randomised, placebo-controlled trial.

RATIONALE: Pulse glucocorticoid therapy is used in hyperinflammation related to coronavirus disease 2019 (COVID-19). We evaluated the efficacy and safety of pulse intravenous methylprednisolone in addition to standard treatment in COVID-19 pneumonia.
METHODS: In this multicentre, randomised, double-blind, placebo-controlled trial, 304 hospitalised patients with COVID-19 pneumonia were randomised to receive 1 g of methylprednisolone intravenously for three consecutive days or placebo in addition to standard dexamethasone. The primary outcome was the duration of patient hospitalisation, calculated as the time interval between randomisation and hospital discharge without the need for supplementary oxygen. The key secondary outcomes were survival free from invasive ventilation with orotracheal intubation and overall survival.
RESULTS: Overall, 112 (75.4%) out of 151 patients in the pulse methylprednisolone arm and 111 (75.2%) of 150 in the placebo arm were discharged from hospital without oxygen within 30 days from randomisation. Median time to discharge was similar in both groups (15 days, 95% CI 13.0-17.0 days and 16 days, 95% CI 13.8-18.2 days, respectively; hazard ratio (HR) 0.92, 95% CI 0.71-1.20; p=0.528). No significant differences between pulse methylprednisolone and placebo arms were observed in terms of admission to intensive care unit with orotracheal intubation or death (20.0% versus 16.1%; HR 1.26, 95% CI 0.74-2.16; p=0.176) or overall mortality (10.0% versus 12.2%; HR 0.83, 95% CI 0.42-1.64; p=0.584). Serious adverse events occurred with similar frequency in the two groups.
CONCLUSIONS: Methylprenisolone pulse therapy added to dexamethasone was not of benefit in patients with COVID-19 pneumonia.

Pharmacologic effective treatments for coronavirus disease 2019 (COVID-19) are needed [1]. Glucocorticoids for their broad and rapid anti-inflammatory effects are ideal candidates for the treatment of hyperinflammation in COVID-19 [2-4]. Interestingly, Fauci and coworkers in the 70 s showed that glucocorticoids block inflammation by inhibiting the efflux of neutrophils and monocytes to the inflammatory sites [5]. Both cell types play an important role in COVID-19-induced lesions, both in the peripheral blood and lungs [3, 6]. The Randomized Evaluation of COVID-19 Therapy (RECOVERY) trial showed that low-dose dexamethasone reduced mortality in hospitalised patients with COVID-19 requiring respiratory support [7]. This observation was confirmed by a prospective meta-analysis of 7 randomised trials in severe COVID-19 patients which showed that the administration of systemic glucocorticoids, compared with usual care or placebo, was associated with lower 28-day all-cause mortality and that it was also safe [8].

Pulse glucocorticoid therapy (>250 mg of prednisone equivalent per day for 1 or a few days) has the most relevant genomic and nongenomic actions, that are responsible for the anti-inflammatory and rapid effects of glucocorticoids, therefore this treatment modality is used for particularly severe and/or life threatening immuno-inflammatory diseases as initial therapy [9-12]. In COVID-19 the cytokine storm represents the acme of the inflammatory process, suggesting the need for a prompt and strong anti-inflammatory effect. Therefore, the addition of pulse glucocorticoid therapy to the standard low dose of dexamethasone scheme can suppresses the hyperinflammatory processes in COVID-19 more effectively than dexamethasone alone and may represent a potential treatment option for patients with severe and critical COVID-19, where pulse glucocorticoid therapy can provide better alternative than non-pulse treatment.

No double-blind randomised trials have been conducted to assess the efficacy of the addition of steroid pulse therapy to the available usual care for COVID-19 pneumonia. To evaluate the efficacy and safety of pulse intravenous methylprednisolone therapy in hospitalised patients with COVID-19 pneumonia, we conducted a multicenter, randomised, double-blind, placebo-controlled trial.

Methods

Study design and participants

This is a multicenter, randomised, double-blind, placebo-controlled trial coordinated by the Azienda Unità Sanitaria Locale-IRCCS of Reggio Emilia, Italy. Nineteen Italian centres participated to this trial.

The trial was submitted and approved on November 20, 2020, by the Italian Medicines Agency (AIFA) (code 130662) and on November 25, 2020, by the COVID-19 Ethics Committee established at the National Institute for Infectious Diseases, Lazzaro Spallanzani, Rome (code 217). The trial was conducted in accordance with the principles of the Good Clinical Practice guidelines of the International Conference on Harmonization. The trial was overseen by an independent data safety and monitoring committee. Written informed consent was obtained from each patient or from the patient's legally authorised representative if the patient was unable to provide consent. Full details of the trial design, conduct, oversight, and analyses can be found in the protocol, statistical analysis plan, and supplementary appendix.

The study population included patients hospitalised for recent-onset COVID-19 pneumonia documented by radiological imaging, with more than five days since initial symptoms of infection, requiring supplemental oxygen in any delivery mode, except invasive mechanical ventilation, with PaO2/FiO2 between 100 and 300, and a C-reactive protein (CRP) greater than 5 mg·dL−1. This CRP value was selected because this cut-off on admission is an indicator of COVID-19 disease severity associated with increased mortality in hospitalised patients [13].

Cases of COVID-19 were confirmed by polymerase chain reaction method with nasopharyngeal swab.

Patients were excluded if they required invasive mechanical ventilation or in the presence of shock or concomitant organ failure requiring an Intensive Care Unit (ICU) admittance. Other exclusion criteria were pregnancy or breastfeeding, severe cardiac or renal failure, diabetes not in good metabolic control based on physician's clinical judgment and other clinical conditions which contraindicate methylprednisolone and which cannot be treated or resolved based on physician's clinical judgment. Therapy with steroid high dose pulses in the week preceding the enrolment in the study and enrolment in another clinical trial were also exclusion criteria.

Randomisation

Eligible patients were randomly assigned in 1:1 ratio to receive either methylprednisolone pulse therapy or placebo, in both arms in addition to standard treatment with dexamethasone. Random assignment was performed centrally at the Trials Center of the Policlinico San Martino-IRCCS in Genova, Italy, stratified by participating centers and type of noninvasive respiratory support [standard oxygen versus High Flow Nasal Cannula (HFNC) or noninvasive ventilation (NIV)]. Computer generated random lists were prepared using permuted balanced blocks of size 4 in random sequence. An Internet-based randomisation system ensured concealment of the treatment assignment until the patient had been registered in the system. Treatment allocation was communicated electronically only to the study center's pharmacy to allow local preparation of the methylprednisolone or placebo bag for the IV administration of the treatment.

Treatment

Patients in the experimental arm received 3 boluses of 1 g of methylprednisolone intravenously on days 1, 2 and 3 from randomisation in 100mL physiological in addition to standard treatment. Patients in the control arm received 100mL of saline (placebo) administered on days 1, 2 and 3 from randomisation in addition to standard treatment. Standard treatment included dexamethasone (6 mg/day oral or intravenous for 10 days), according to the RECOVERY study protocol [2]. The standard therapy was allowed to change during the study in accordance with the indications of the Italian Medicines Agency (AIFA) cards regarding the choice of drugs, dosages, or duration of treatment.

Procedures

Patients were evaluated daily during their hospitalisation, from randomisation (day 1) through day 30. The type of noninvasive respiratory support was daily recorded. The medical evaluations on day 21, 28 and end of study (30 days) could be limited to a telephone contact if the patient was at home. The compilation of the end of study visit was mandatory and it could coincide with any clinical visit or telephone contact occurred during the 30 days’ observation window. All the severe and non-severe adverse events observed during the hospitalisation and in the 30 days from randomisation were recorded in the clinical database. Adverse events were classified and graded according to the Common Terminology Criteria for Adverse Events (CTCAE) v 5.0 criteria. For each adverse event the maximum grade of toxicity was identified. Trial participants, investigators, study team care providers were unaware of the trial-group assignments until after all data queries were resolved and the database was locked. Only the hospital pharmacies were aware of the treatment allocation.

Outcomes

The primary endpoint was the duration of the patient hospitalisation, calculated as the interval of time between the randomisation and the hospital discharge without the need of supplementary oxygen during the first 30 days following randomisation. The patients who were transferred to other hospital wards at different intensity of care were considered as still hospitalised. Patients who were transferred to other institutions outside the hospital (e.g., COVID-hotels) for the impossibility of transferring the patient at home were considered as discharged from the hospital.

Patients deceased within 30 days were considered as never discharged.

Patients discharged at home with supplementary oxygen were censored at the date of discharge because no further follow-up information was available. This choice implied the questionable assumption that their prognosis was similar to that of patients still in hospital: the effects of this assumption were evaluated in 2 post-hoc sensitivity analyses in which these patients were considered either discharged without oxygen, or never discharged. The results of these analyses closely resembled those of the primary analysis. The secondary endpoints of efficacy were the following: According to the study protocol, the following explorative endpoints were also evaluated in selected populations:

Survival free from invasive ventilation with orotracheal intubation, defined as the interval between the randomisation and the first use of invasive ventilation with orotracheal intubation or death.

Overall survival, defined as the interval between randomisation and death for any cause.

Survival free from clinical worsening, defined as the interval between randomisation and the first episode in which the ratio PaO2/FiO2 drops below the value 150 or death. Only patients with PaO2/FiO2, >200 mmHg at enrolment were included in this analysis.

Survival free from NIV or HFNC, defined as the interval between randomisation and the first use of NIV or the first administration of HFNC, or death. Only patients who did not receive oxygen in this modality at randomisation were included in this analysis.

Statistical analysis

Assuming a median time to hospital discharge of 14 days in the control arm, the experimental treatment was considered effective if it reduced time to discharge by 37.5% in relative terms (hazard ratio=1.60). This corresponds to an absolute increase in the cumulative probability of discharge at 14 days from 50% to 67%, i.e., a decrease in the median time to discharge from 14 to less than 9 days. Assuming a statistical power of 90% and a 5% error rate α, for a two-tailed log rank test at least 198 hospital discharges needed to be observed. The sample size required to observe 198 hospital discharges, assuming a 5% overall dropout rate, was estimated to be at least 260 patients, 130 per arm.

Statistical analysis is detailed in the Statistical Analysis Plan. The cumulative probabilities of survival free from discharge in the two treatment groups were estimated according to the Kaplan Meier method and compared with a non-stratified log rank test. The hazard ratio and its 95% confidence interval (95%CI) were computed by fitting a univariate Cox's model. All secondary efficacy analyses were performed with the same approach described for the primary efficacy analysis. Primary and secondary efficacy analyses were performed on all patients included in the intention-to-treat population. Explorative endpoints analyses were carried out on the appropriate sub-populations, identified from the intention-to-treat population.

We reported the p value (2 tailed; significance defined as p<0.05) and bilateral 95%CI not adjusted for multiplicity.

We performed some subgroups analyses not planned into the Statistical Analysis Plan that we reported into the Appendix. The subgroups were defined according to the PaO2/FiO2 values at randomisation (PaO2/FiO2, between 100 and 200 mmHg versus PaO2/FiO2 between 201 and 300 mmHg), the modality of oxygen administration (supplemental standard oxygen versus noninvasive ventilation or high-flow oxygen), CRP values (stratified in >5 and ≤10 mg·dL−1, >10 and ≤15 mg·dL−1, and >15 mg·dL−1) at randomisation, age (≤60 years and >60 years), and Body Mass Index (BMI) (stratified in quartiles: 1st quartile: ≤24.6; 2nd quartile: 24.7–27.6; 3rd quartile: 27.7–30.9; 4th quartile: ≥31.0). Hazard ratios estimates obtained in the Cox model were reported within each subgroup. The presence of a significant variation of the Hazard Ratios across strata of each factor was assessed by means of the treatment-by-factor interaction tests.

Statistical analyses were performed using SAS, version 9.4 (SAS Institute) and SPSS, version 23 (IBM Corp). The trial is registered with EudraCT, 2020-004323-16, and ClinicalTrials.gov Identifier, NCT04673162.

Results

Patients

Between December 21, 2020, and March 10, 2021, the 19 centres randomised 304 patients, 152 in the methylprednisolone pulses arm and 152 in the placebo and standard therapy arm. Three patients (1 in the experimental arm and 2 in the control arm) withdrew consent leaving 301 patients eligible for the intention-to-treat analysis (fig. 1).

FIGURE 1

The CONSORT Flow Diagram of the study.

One hundred and forty-nine patients (98.7%) assigned to receive methylprednisolone pulses and 148 (98.7%) assigned to receive placebo received the therapy as assigned. Most patients, 144/151 (95.4%) in active treatment arm and 142/150 (94.7%) in the placebo arm, received all 3 pulses. Demographic and baseline clinical characteristics are summarised in table 1.

TABLE 1

Demographic and clinical characteristics of patients at randomization

Characteristics	Placebo+Standard (N=150)	MTP pulses+Standard (N=151)
Age, median (IQR) – y	64.0 (55.0–72.2)	64.0 (54.0–74.0)
Male sex – no. (%)	106 (70.7)	111 (73.5)
Days from symptom onset to randomisation, median (IQR)	8.0 (6.0–10.0)	9.0 (6.0–11.0)
Days from Pneumonia diagnosis to randomisation, median (IQR)	1.0 (1.0–2.0)	1.0 (1.0–2.0)
PaO2/FiO2, median (IQR) – mmHg	204.0 (158.0–243.0)	208.0 (158.0–248.0)
Respiratory rate, median (IQR) – (breaths/min)	20.0 (18.0–24.0)	20.5 (18.0–24.0)
CRP, median (IQR) – mg·dL−1	10.9 (7.6–14.6)	10.6 (7.3–14.9)
Modality of oxygen administration (%)
Supplemental standard oxygen	101 (67.3)	104 (68.9)
NIV or high-flow oxygen	49 (32.7)	47 (31.1)
Coexisting conditions – no. (%)
Diabetes mellitus	19 (12.7)	26 (17.2)
Hypertension	74 (49.3)	83 (55.0)
COPD	8 (5.3)	5 (3.3)
Heart failure	7 (4.7)	9 (6.0)
Obesity (BMI≥30 kg·m−2)	35/112 (31.3)	34/121 (28.1)
Other Treatments – no. (%) #
Glucocorticoids	135 (90.0)	131 (86.8)
Remdesivir	27 (18.0)	19 (12.6)
Antibiotics	41 (27.3)	40 (26.5)
LMWH	117 (78.0)	120 (79.5)

MTP: methylprednisolone; IQR: interquartile range; CRP: C-reactive protein; NIV: non-invasive ventilation; COPD: Chronic obstructive pulmonary disease; LMWH: Low-molecular-weight heparin.

#Any treatments administered during the period from the onset of symptoms until randomisation.

Primary outcome

One hundred and twelve of 151 (75.4%) patients in the pulse methylprednisolone arm and 111 of 150 (75.2%) patients in the placebo arm were discharged from hospital without oxygen within 30 days. No difference in time to discharge without oxygen was observed between the two groups: the median was 15 (95%CI, 13.0 to 17.0) days in methylprednisolone arm versus 16 (95%CI, 13.8 to 18.2) days in placebo arm (hazard ratio, 0.92; 95%CI, 0.71–1.20; p-value= 0.528) (table 2 and fig. 2a).

TABLE 2

Clinical outcomes in the intention-to-treat population

Endpoints	Placebo+Standard	MTP pulses+Standard	Hazard Ratio (95% CI)	p-value
	no. (% at 30 days)	no. (% at 30 days)
Primary endpoint
Discharge without oxygen#	111/150 (75.2)	112/151 (75.4)	0.92 (0.71–1.20)	0.528
Time to discharge within 30 days – median (95%CI)	16 (13.8–18.2)	15 (13.0–17.0)
Secondary endpoints
Admission to ICU or death¶	24/150 (16.1)	30/151 (20.0)	1.26 (0.74–2.16)	0.176
Deaths	18/150 (12.2)	15/151 (10.0)	0.83 (0.42–1.64)	0.584
Explorative endpoints
Clinical worsening or death+	31/77 (40.3)	37/81 (45.7)	1.17 (0.73–1.89)	0.430
Use of HF, NIV, or death§	51/101 (50.6)	55/104 (52.9)	1.16 (0.79–1.70)	0.430

All proportions were cumulative probabilities estimated with the Kaplan Meier Product Limit Estimator and compared with the log-rank test.

ICU: Intensive Care Unit; HF: High Flow oxygen; NIV: non-invasive ventilation.

#1 patient in the Placebo arm and 3 patients in the MTP arm had been admitted to ICU but had a full recovery and were discharged without oxygen within 30 days.

¶8 patients in the Placebo arm and 7 patients in the MTP arm died after admission to ICU, while 10 patients and 8 patients, respectively, died before being admitted to ICU.

+In the 158 patients with PaO2/FIO2 ratio>200 mmHg at randomisation.

§In the 205 patients who received standard supplementary oxygen at randomisation.

FIGURE 2

Kaplan-Meier estimates of hospital discharge without oxygen (a) orotracheal intubation or death (b), and death (c).

Secondary outcomes

Fifty-four patients were admitted to Intensive Care Unit and received invasive ventilation with orotracheal intubation or died, with no significant differences between pulse methylprednisolone and placebo arms (20.0% versus 16.1%; hazard ratio, 1.26; 95%CI, 0.74–2.16) (table 2 and fig. 2b).

Thirty-three deaths occurred within 30 days since randomisation, and mortality was comparable in the two arms (10.0% versus 12.2%; hazard ratio, 0.83; 95%CI, 0.42–1.64) (table 2 and fig. 2c).

Explorative endpoints and subgroup analyses

At randomisation, 158 patients had PaO2/FiO2 ratio >200 mmHg. In these patients the incidence of clinical worsening defined as progression to a PaO2/FiO2<150 mmHg or death was similar in methylprednisolone pulse and placebo arms (45.7% versus 40.3%; hazard ratio, 1.17; 95%CI, 0.73–1.89) (table 2).

At randomisation, 205 patients received standard supplementary oxygen. In these patients the incidence of progression to the use of HFNC or NIV, mechanical ventilation or death was not different in methylprednisolone pulse and placebo arms (52.9% versus 50.6%; hazard ratio, 1.16; 95%CI, 0.79–1.70).

No differences were observed between the two arms in subgroup analyses, where subgroups were defined according to PaO2/FiO2 values, different modalities of oxygen therapy, CRP values, age and BMI at randomisation (see Appendix).

Safety outcomes

Eighty-six adverse events were reported in 51/149 (34.2%) patients treated with placebo and 90 in 54/149 (36.2%) patients treated with pulses of methylprednisolone. Twenty-eight (32.6%) grade 3 or 4 adverse events occurred in placebo group versus 35 (38.9%) in pulse methylprednisolone group (table 3). Forty-eight adverse events were judged to be treatment related by the principal investigators: 18 (20.9%) and 30 (33.3%) in the placebo and methylprednisolone arm, respectively. Adverse events by MedDRA system organ class (version 22.0) are listed in Appendix.

TABLE 3

Adverse events by treatment arm, grade and relatedness in the safety population of 297 patients

	Treatment Arm
	Placebo+Standard	MTP pulses+Standard
Adverse Events – no. (%)
Reported AEs	86	90
Grade AEs#
1–2	51 (59.3)	52 (57.8)
3–4	28 (32.6)	35 (38.9)
5	7 (8.1)	3 (3.3)
Relatedness AEs (Yes)	18 (20.9)	30 (33.3)
Patients with AEs	51/149 (34.2)	54/149 (36.2)
Serious Adverse Events – no. (%)
Reported SAEs	16	9
Grade SAEs#
1–2	1 (6.2)	–
3–4	8 (50.0)	7 (77.8)
5	7 (43.8)	2 (22.2)
Relatedness SAEs (Yes)	3 (18.7)	2 (22.2)
Patients with SAEs	12/149 (8.0)	8/149 (5.4)

#Grade AEs: 1=Mild; 2=Moderate; 3=Severe; 4=Life-Threatening; 5=Death.

Sixteen serious adverse events (SAEs) occurred in the placebo arm and 9 in the methylprednisolone pulses. Eight grade 3 or 4 SAEs occurred in the placebo arm and 7 in methylprednisolone pulses. The number of SAEs considered by the investigators as treatment related were similar in the two arms, as well as the frequency of patients with SAEs (table 3). A summary description of SAEs is reported in the Appendix. Bacterial infections were the most frequently reported severe adverse events in both arms and they were detailed in table S2. In particular, serious infections thought to be related to the treatment were observed in 3.3% of the patients treated with methylprednisolone pulses and in 4% of those treated with placebo. Cases of severe uncontrolled hyperglycemia were not reported in the two arms.

Discussion

In this double-blind, randomised, placebo-controlled trial no significant difference was observed in time to hospital discharge between the high-dose methylprednisolone pulse group and the standard of care group in patients with COVID-19 pneumonia. There was also no benefit in the secondary outcomes, survival and admission to intensive care unit. The lack of benefit on the primary outcome detected in the ITT population was also observed in subgroups of patients defined according to different PaO2/FiO2 values, different modalities of oxygen therapy, and different CRP values at randomisation. Furthermore, we failed to observe beneficial effects of methylprednisolone pulses on progression of respiratory failure in the subgroups of 158 patients with PaO2/FiO2 ratio >200 mmHg and in the 206 patients who received standard supplementary oxygen at randomisation. Therefore, a rapid and vigorous suppression of inflammation with the addition of methylprednisolone pulses did not provide clinical benefit compared to standard care with dexamethasone according to the RECOVERY trial schedule in COVID-19 hospitalised patients requiring oxygen therapy [7].

Pulse glucocorticoid therapy is used in many immuno-inflammatory conditions to obtain a quick and strong suppression of inflammation in emergency situations [9–12, 14]. However, despite its diffuse use in clinical practice, very few studies, often uncontrolled and underpowered, have evaluated the efficacy and safety of this therapy [10–12, 14–17]. Indeed, a systematic literature review on safety and efficacy of pulse glucocorticoid therapy for SARS-CoV, Middle East Respiratory Syndrome (MERS)-CoV or SARS-CoV-2 showed that the quality of the evidence is poor and randomised controlled trials are highly needed [14]. A recent Iranian randomised trial showed a significant pulmonary improvement and a reduced mortality in hospitalised patients with severe COVID-19 treated with pulse methylprednisolone compared to those treated with standard therapy [18]. However, this study has several limitations. It was not double-blind placebo controlled and enrolled only 62 patients. Even more important, differently from our study, standard of care did not include glucocorticoid treatment. A Spanish randomised, open-label, controlled study in hospitalised patients with COVID-19 pneumonia needing oxygen therapy compared low dose dexamethasone (6 mg once daily for 10 days) to high dose dexamethasone (20 mg once daily for 5 days, followed by 10 mg once daily for additional 5 days) [19]. Similarly to our study, this trial showed no difference in efficacy between high dose and low dose of dexamethasone. However, the two studies were not completely comparable because we used in pulse therapy a 10 times higher steroid dosage (125 mg of prednisone equivalent per day versus 1250 mg of prednisone equivalent per day), maximising the corticosteroid genomic and nongenomic actions [9]. Although all our patients had severe inflammation at enrollment (CRP values had to be more than 10 times the upper reference), they represented a spectrum of COVID-19 disease, ranging from patients more inflamed and critically ill at the time of randomisation to those with less severe inflammatory disease receiving supplemental oxygen by nasal prongs. We only excluded patients in early symptomatic phases (i.e., interval between initial symptoms of infection and randomisation ≤5 days) as we did not want to prevent functional immune response in the early stages of the infection, and patients in invasive mechanical ventilation or with presence of shock or concomitant organ failure requiring an Intensive Care Unit admittance.

Despite the overall negative results of our study, we cannot exclude potential benefit in specific subgroups of patients with more severe disease and more severe inflammatory response. Also, our study cannot rule out a role for pulse glucocorticoid therapy as a rescue therapy in patients who failed to improve after dexamethasone and that a gradual tapering of pulse methylprednisolone would have better controlled the inflammation, avoiding possible rebound. Of interest, our study did not show an increased risk of glucocorticoid side effects, particularly infections or hyperglycemia, in the patients treated with methylprednisolone pulses. Previous studies observed only an increased frequency of mild acute adverse events (e.g., sleep disturbance, mood change) in patients treated with pulse glucocorticoids compared to those treated with placebo [10, 12]. Therefore, our results confirm that pulses of methylprednisolone are safe also in the inflammatory phase of patients with COVID-19.

Our trial has some strengths, but also some limitations. One of the strengths is that the trial was multicenter double-blind placebo controlled. Our trial population was intentionally chosen to have a severe inflammatory status as reflected by the elevated CRP values at baseline which were similar in the pulse methylprednisolone and placebo patients. These patients were presumably those with better chance of responding to methylprednisolone pulses. The two groups were balanced in terms of their baseline characteristics, including demographics, days from symptom onset to randomisation, PaO2/FiO2 value and modality of noninvasive support at randomisation, coexisting conditions, and concurrent treatment. The selected primary outcome, duration of patient hospitalisation, may have important limitations, particularly regarding differences in local clinical practice. However, in Italy, during the study period, the treatment of hospitalised patients with COVID-19 was standardised by AIFA and the treatments were similar across all trial sites.

The main limitation of this trial derived from its limited sample size, which implies that the observed results, do not allow to rule out the hypothesis that pulses of methylprednisolone are associated with a modest beneficial effect, e.g., a 15–16% reduction in time to discharge. More important, we cannot exclude possible benefits in specific subpopulations of patients that were underpowered to detect clinically relevant differences. Platform trials enrolling large numbers of patients could better evaluate the efficacy of pulse glucocorticoid therapy in these specific subgroups of patients.