Short-term inhalation of sargramostim with concomitant high-dose steroids does not hasten recovery in moderate COVID-19 pneumonia: a double-blind, randomised, placebo-controlled trial

Abstract Background Granulocyte-macrophage colony stimulating factor (GM-CSF) inhalation may alleviate pulmonary inflammation caused by viral pneumonia. To investigate this, we evaluated its efficacy on COVID-19 pneumonia. Methods This double-blind, randomised, placebo-controlled study (ClinicalTrials.gov: NCT04642950) evaluated patients in the first half of 2021 at seven Japanese hospitals. Hospitalised patients with COVID-19 pneumonia with moderate hypoxaemia inhaled sargramostim or placebo for 5 days. The primary endpoint was days to achieve a ≥ 2-category improvement from baseline on a modified 7-category ordinal scale. Secondary endpoints included degree of oxygenation, defined by amount of oxygen supply, and serum CCL17 level. Results Seventy-five patients were randomly assigned in a 2:1 ratio to receive sargramostim or placebo, of which 47 and 23 were analysed, respectively. No difference was observed between groups regarding the primary endpoint (8.0 and 7.0 days for sargramostim and placebo, respectively) or in the secondary endpoints, except for CCL17. A post hoc sub-analysis indicated that endpoint assessments were influenced by concomitant corticosteroid therapy. When the cumulative corticosteroid dose was ≤500 mg during Days 1–5, recovery and oxygenation were faster in the sargramostim group than for placebo. Bolus dose corticosteroids were associated with temporarily impaired oxygenation and delayed clinical recovery. The increase in serum CCL17, a candidate prognostic factor, reflected improvement with sargramostim inhalation. The number of adverse events was similar between groups. Two serious adverse events were observed in the sargramostim group without causal relation. Conclusions Inhaled sargramostim was likely to be effective for COVID-19 pneumonia unless the concomitant corticosteroid dose was high.


Introduction
The ongoing coronavirus disease 2019 (COVID-19) pandemic, which is caused by the SARS-CoV-2 virus, has resulted in over 767 million confirmed cases of COVID-19 globally and more than 33 million cases in Japan, with the number of deaths reaching over 6 million worldwide [1,2].The impact of the pandemic, including the crisis of hospital bed shortages, has decreased due to the widespread use of vaccines, with over 13 billion vaccine doses being administered worldwide [1].However, the emergence of new variants has led to repeated waves of infection.
Patients infected with SARS-CoV-2 can experience variable symptoms that can be categorised as mild, moderate I, moderate II, or severe [3], with reduced oxygen saturation (SpO 2 ) being associated with increasingly serious clinical symptoms.In mild, moderate I, and moderate II cases, those without pneumonia are associated with SpO 2 >96%, while cases with pneumonia have SpO 2 between 94%-95%, and those with SpO 2 <93% have pneumonia and require oxygen supplementation.If respiratory failure still progresses, the patient is considered to have severe disease and may require admission to an intensive care unit (ICU) and/or mechanical ventilatory support [3].In such severe cases, COVID-19 pneumonia progresses to acute respiratory distress syndrome (ARDS), which may lead to multi-organ failure and death [4].Cytokine storms may contribute to the progression from pneumonia to ARDS [5,6].Evidence of pulmonary inflammation and immunopathology without detectable virus can be found in autopsies of COVID-19 patients [7], highlighting the central role that the aberrant immune response plays in mortality.
Previous research has shown that corticosteroids, such as dexamethasone, may be effective in preventing progression from pneumonia to ARDS [8].Corticosteroid therapy is widely used to treat patients with the moderate disease [9,10]; however, the use of high-dose rescue boluses of corticosteroids in patients with refractory hypoxaemia has been reported to be associated with higher rates of admissions to ICU and in-hospital mortality, longer durations of ventilator assistance, and an increased prevalence of nosocomial infection [11].Thus, the use of high-dose corticosteroid therapy is controversial [11,12].
Granulocyte-macrophage colony-stimulating factor (GM-CSF) is a crucial factor that promotes the terminal differentiation of blood-derived monocytes to mature alveolar macrophages, which not only plays an important role in cleaning the lower respiratory tract, but also in suppressing inflammation in situ [13].The activation of this pathway and the subsequent increase in mature alveolar macrophages has been shown to reduce inflammation in the lungs of mice [13].When monocytes are instead stimulated by inflammatory cytokines, such as interleukin (IL)-6 or IL-1b, they differentiate into alternatively activated macrophages, which have been shown to increase the severity and progression to fibrotic lung disease in a mouse model of SARS-CoV-2 infection [14].However, when GM-CSF is either administered or expressed locally in the lungs, the alternatively activated macrophages differentiate into mature alveolar macrophages [13].Thus, GM-CSF is an effective treatment for both bacterial and viral pneumonias and associated inflammation, and has been successfully used to treat pneumonia-associated ARDS in humans [15].Moreover, inhaled GM-CSF is reported to enhance anti-viral T and B cell effectors recognising SARS-CoV-2-infected cells [16].
Among the physicians who treated patients with moderate COVID-19 pneumonia in Japan, some prescribed high-dose corticosteroids instead of the National Institutes of Health (NIH)-recommended regimen [17] with the goal of improving the pneumonia as early as possible.Therefore, in this study, we allowed the site pulmonologist to choose a corticosteroid combination therapy, resulting in half of the participants receiving high-dose corticosteroid therapy.However, owing to this, we were able to investigate the relationship between steroid dose and response to GM-CSF inhalation therapy.

Study design
This was a double-blind, randomised, placebo-controlled, multi-centre study (ClinicalTrials.govidentifier: NCT04642950), in which patients were enrolled between February 8 th and October 25 th , 2021 in Japan.Patients were treated with the study drug on an inpatient basis for 5-10 days and were monitored for up to 28 days.Approved concomitant medications were corticosteroids, remdesivir, and baricitinib.
The study met all legal and regulatory requirements and was conducted in accordance with the ethical principles laid out in the Declaration of Helsinki and the Pharmaceutical and Medical Devices Law in Japan.The institutional review board of each participating centre approved the study protocol and all related documentation.All patients provided written informed consent before entering the study.

Participants
The participants in this study were hospitalised Japanese patients aged 20-84 years who tested positive for SARS-CoV-2 infection confirmed by polymerase chain reaction test, with clinically diagnosed pneumonia and with moderate hypoxaemia (SpO 2 93%) when breathing room air in the supine position.The exclusion criteria are listed in the supplementary materials (Text S1).

Randomisation and masking
Patients were randomly assigned in a 2:1 ratio to receive either sargramostim or placebo inhalation, respectively.A web-based randomisation system used minimisation to ensure a balanced allocation of patients receiving concomitant remdesivir to each group.Randomisation was performed by an unmasked pharmacist, with the allocation information disclosed only to the unmasked monitor.Masking was maintained by the absence of the investigators and study collaborators (except for the unmasked pharmacist) during both the preparation and dispensing of the study drug.

Procedures
The study drug was prepared as 250 mg of sargramostim dissolved in 4 mL of physiological saline, which was administered twice daily as two 125-mg doses via a jet nebuliser (LC Sprint Star Pari, Germany).Treatment continued for at least 5 days and up to 10 days at the discretion of the investigators.The placebo group received physiological saline, administered in the same manner as sargramostim.
The observation and examination schedule throughout the study is provided in Table S1.Patients were monitored by laboratory tests, vital sign assessments, and 12-lead electrocardiography.
In principle, high-resolution computed tomography (HRCT) was to be conducted on every patient; however, chest CT or X-rays were used when HRCT was not possible.Evaluation of HRCT images was conducted by CT densitometry techniques using the SYNAPSE VINCENT volume analyser (Fuji Film) to calculate 'the percent volume', which was defined as the percentage of the lung volume of the HRCT image corresponding to a specific CT density range to the total lung volume of the whole CT density range (minimum to maximum CT value automatically extracted by SYNAPSE VINCENT).Visual scoring of HRCT images was performed by two experienced image readers.The detailed evaluation procedures are described in Text S2.

Outcomes
The primary endpoint was the number of days to achieve clinical improvement, which was defined as a !2-category improvement on a modified 7-category ordinal scale (OS) from baseline to Day 28.The modified 7-category ordinal scale was scored into Categories 1-7, as follows: 1 ¼ patient death; 2 ¼ use of invasive ventilation or extracorporeal membrane oxygenation; 3 ¼ use of non-invasive ventilation/high-flow oxygen supply device; 4 ¼ supplemental oxygen required; 5 ¼ no supplemental oxygen required, but continuous treatment required (for COVID-19 or other conditions); 6 ¼ without supplemental oxygen and the dose of corticosteroid was tapered to 25 mg prednisolone-equivalent; and 7 ¼ patient discharged from hospital.As four of the seven centres used corticosteroid pulse therapy with bolus administration of high-dose corticosteroids, some patients continued tapering even after their discharge from the hospital, and sometimes the order of transition from Category 5 to Category 6 or 7 was reversed.
Therefore, before the key open, the time point of Category 6 was corrected to when oxygen therapy was completed and the dose of corticosteroid was tapered to 25 mg prednisolone-equivalent.Thus, the modified 7-category ordinal scale was applied with this correction.
Secondary endpoints included changes in inflammatory markers and cytokines, changes in pulmonary inflammatory findings, and oxygenation disorder.Oxygenation disorder was evaluated by the degree of oxygenation (DOO) using an ordinal scale of oxygen requirement as follows: 1 ¼ requiring supplemental oxygen delivered at !4 L/min using an oxygen mask or nasal high flow; 2 ¼ requiring supplemental oxygen at !2 L/min using a nasal cannula; 3 ¼ requiring supplemental oxygen at <2 L/min using a nasal cannula; 4 ¼ no supplemental oxygen required, but the patient reports shortness of breath when walking or moving; and 5 ¼ no supplemental oxygen required and no shortness of breath when walking or moving.See Text S3 for full details of the secondary endpoints.Safety endpoints included the incidence of adverse events (AEs) and serious AEs (SAEs), including adverse drug reactions (ADRs).AEs and ADRs were classified according to the Medical Dictionary for Regulatory Activities by System Organ Class and Preferred Term (version 23.0).

Statistical analyses
The methods for determining the target sample size are described in Text S4.The primary analysis was a logrank test of time to improvement with sargramostim compared with placebo in the efficacy analysis set (i.e.patients who received at least one dose of the study drug and whose efficacy data were available).In the analysis of the period to !2-category improvement (primary endpoint), patients who discontinued early and those who did not improve were censored at Day 29, in line with the pre-specified analysis plan.These patients were included in the analysis as those who did not improve until Day 29.For other endpoints, including the post hoc subgroup analyses, the data were summarised by groups and timepoints.In cases of missing data, data were imputed using the last observation carried forward method; this was performed only in cases where the oxygen requirement and modified 7-category ordinal scale data were summarised by timepoint.We also performed confounding analyses for serum CCL17 levels and HRCT; details are described in Text S5.
All statistical analyses had a two-sided significance level of 5%, with a two-sided 95% confidence coefficient (1 À a) applied for interval estimation.Statistical analyses were conducted using SAS version 9.4 (SAS Institute; Cary, NC, USA).

Patients
Of the 76 patients who gave consent, 75 were randomly assigned to either the sargramostim group (50 patients) or the placebo group (25 patients) in a 2:1 ratio (Figure 1).Three of the 50 patients in the sargramostim group were excluded due to investigator's decision by oxygenation impairment (n ¼ 2) and ICU admission (n ¼ 1); 47 were treated.Subsequently, three patients discontinued the study, and 44 completed the 28-day follow-up.No patients died, but two were intubated (one each on Days 1 and 3), and therefore discontinued inhalation treatment.Forty-one patients in the sargramostim group were discharged during the 28-day study period.Two of the 25 patients assigned to placebo were excluded due to investigator's decision by oxygenation impairment and ICU admission (one patient each), and 23 patients completed 28 days of follow-up.No patients in this group died, and 22 were discharged during the study period.The median days to hospital discharge were 11 days for the sargramostim group and 10 days for the placebo group, with a minimum of 6 days for both.Baseline patient demographic and clinical characteristics are shown in Table 1.There were no significant differences in characteristics between the sargramostim and placebo groups.Overall, most patients were male (74.3%), with a mean age of 54.7 years, and a mean BMI of 25.9 kg/m 2 .The mean baseline SpO 2 was 90.6% and 91.3% for the sargramostim and placebo groups, respectively.Overall, most patients were Category 4 on the modified 7-category ordinal scale, and the most common DOO was Category 2.

Primary and secondary endpoints
The median days to clinical improvement were 9.0 days and 8.0 days for the sargramostim (n ¼ 47) and placebo groups (n ¼ 23), respectively, which were not statistically different (log-rank, p ¼ 0.10).Similarly, there was no between-group difference with Category 6 correction (8.0 days in the sargramostim group and 7.0 days in the placebo group; log-rank, p ¼ 0.291; Figure 2).DOO was also not significantly improved in the sargramostim group compared with the placebo group (Figure 3).There were no significant differences between the two groups for the other secondary endpoints (data not shown), except for C-C motif chemokine ligand 17 (CCL17).

Dose distribution of concomitant corticosteroid therapy
At the planning stage of the statistical analysis, we did not anticipate such large doses of steroids to be used in combination with sargramostim inhalation.However, as six patients received cumulative corticosteroid doses (CCDs) of >2500 mg on Days 1-5, and because the effect of highdose corticosteroid therapy on COVID-19 pneumonia might delay clinical improvement [18,19], we decided to perform a post hoc sub-analysis on the modified 7-category ordinal scale and DOO according to the CCD on Days 1-5.All except two patients (n ¼ 1 in each group) received corticosteroids, but the type and dose varied.As 1 mg of dexamethasone and methylprednisolone are known to be equivalent to 6.67 mg and 1.25 mg of prednisolone, respectively, the CCD during the study drug inhalation period, i.e.Days 1-5, varied greatly, as shown in Table 2 and Figure S1.Twenty-seven patients (57.4%) and 16 patients (69.6%) in the sargramostim and placebo groups, respectively, received a CCD of 500 mg.Of these patients, 37 received dexamethasone (n ¼ 35) or no corticosteroids (n ¼ 2), and six received methylprednisolone.Fourteen patients (29.8%) in the sargramostim group and seven (30.4%) in the placebo group received a CCD of >500-2500 mg, given as methylprednisolone only.Furthermore, six patients (12.8%) in the sargramostim group received a >2500-mg CCD as bolus methylprednisolone, which was mainly administered at three centres (p ¼ 0.002, n ¼ 33).

Effect of corticosteroid dose on clinical improvement and DOO
As a post hoc analysis, we investigated the effect of corticosteroid dose on oxygenation and clinical improvement regardless of sargramostim or placebo inhalation.A generalised estimating equation suggested that the CCD on Days 1-5 affected DOO until Day 5. Patients with a CCD >500 mg had a delayed improvement in DOO compared to those with a CCD 500 mg (Figure 4A).In particular, in all six patients who received a CCD >2500 mg, a transient worsening of DOO (early dip) was observed by Day 5.One patient was intubated, starting ventilation on Day 3. As shown in Figure 4B, the number of days to clinical improvement varied according to CCD ( 500 mg, >500-2500 mg, and >2500 mg were given on Days 7, 9, and 13, respectively; p ¼ 0.002).

Effect of sargramostim inhalation on improvement in oxygenation by corticosteroid dose
Next, we examined the effects of sargramostim inhalation on oxygenation and clinical improvement, taking corticosteroid dose into account as a post hoc analysis.Figure 5 shows the number of days to clinical improvement, comparing between the treatment groups and between subgroups with a CCD 500 mg and >500 mg (Figure 5A, B) on Days 1-5.The median number of days to clinical improvement in patients receiving a 500 mg CCD was 6 days for the sargramostim group and 7.5 days for the placebo group.In contrast, the median number of days to clinical improvement in patients receiving a >500 mg CCD was 10.5 days in the sargramostim group and 7 days in the placebo group.There was a significant difference among the four groups in the number of days until clinical improvement (p ¼ 0.007), suggesting that the corticosteroid dose was related to clinical improvement by sargramostim inhalation.Improvement in DOO by CCD ( 500 mg, >500 mg; Figure 5C, D) and by administration group on Days 1-5 was faster in the sargramostim group than the placebo group for patients receiving a 500 mg CCD; this effect was reversed at >500 mg.Given the negative impact of high doses of corticosteroids in the placebo arm and the improvement of oxygenation in the sargramostim b DexN was defined as standard treatment plus dexamethasone or standard treatment alone.mPSL was defined as standard treatment plus methylprednisolone or prednisolone at least once.c SpO 2 was measured by having the patients rest in the supine position on a bed in the medical or laboratory room for at least 5 min, breathing room air (the oxygen supply device was temporarily removed during inhalation), and confirming that SpO 2 was stable with a pulse oximeter.d The modified 7-category ordinal scale was defined as follows: 1 ¼ patient death; 2 ¼ use of invasive ventilation or extracorporeal membrane oxygenation; 3 ¼ use of non-invasive ventilation/high-flow oxygen supply device; 4 ¼ supplemental oxygen required; 5 ¼ no supplemental oxygen required, but continuous treatment required (for COVID-19 or other conditions); 6 ¼ without supplemental oxygen and the dose of corticosteroid was tapered to 25 mg prednisolone-equivalent; 7 ¼ patient discharged from hospital.¶ Degree of oxygenation was defined as follows: 1 ¼ requiring supplemental oxygen delivered at !4 L/min using an oxygen mask or nasal high flow; 2 ¼ requiring supplemental oxygen at !2 L/min using a nasal cannula; 3 ¼ requiring supplemental oxygen at <2 L/min using a nasal cannula; 4 ¼ no supplemental oxygen required, but the patient reports shortness of breath when walking or moving; 5 ¼ no supplemental oxygen required and no shortness of breath when walking or moving.JAK: Janus kinase; SD: standard deviation; SpO 2 : oxygen saturation.Kaplan-Meier curves showing the percentage of patients (vertical axis) who achieved improvement in two or more categories on the modified 7-category ordinal scale from baseline over days since enrolment (horizontal axis) in the sargramostim (blue line) and placebo (red line) groups.There was no difference between the sargramostim and placebo groups.Modified 7-category ordinal scale: 1 ¼ patient death; 2 ¼ use of invasive ventilation or extracorporeal membrane oxygenation; 3 ¼ use of non-invasive ventilation/high-flow oxygen supply device; 4 ¼ supplemental oxygen required; 5 ¼ no supplemental oxygen required, but continuous treatment required (for COVID-19 or other conditions); 6 ¼ without supplemental oxygen and the dose of corticosteroid was tapered to 25 mg prednisolone-equivalent; and 7 ¼ patient discharged from hospital.The degree of oxygenation in the sargramostim group did not differ from the placebo group.Degree of oxygenation: 1 ¼ requiring supplemental oxygen delivered at !4 L/min using an oxygen mask or nasal high flow; 2 ¼ requiring supplemental oxygen at !2 L/min using a nasal cannula; 3 ¼ requiring supplemental oxygen at <2 L/min using a nasal cannula; 4 ¼ no supplemental oxygen required, but the patient reports shortness of breath when walking or moving; 5 ¼ no supplemental oxygen required and no shortness of breath when walking or moving.
arm on the lower steroid arm, these results suggest that corticosteroid dose affects the efficacy of sargramostim on oxygenation and clinical improvement.

Relationship between CCL17 increase and clinical improvement/DOO
According to the statistical plan, serum concentrations of IL-6, IL-1b, tumour necrosis factor-a, IL-10, CCL17, chemokine ligand 9, and matrix metalloproteinase-12 were serially measured and compared between sargramostim and placebo groups.Only CCL17 was significantly increased in the sargramostim group compared with the placebo group (Figure 6A, Day 5: p < 0.001).As a post hoc analysis, when the change in serum CCL17 (DCCL17) on Days 1-5 was stratified into groups based on the median value (high DCCL17 group, !1277 pg/mL; low DCCL17 group, <1277 pg/mL), we found that all patients in the high DCCL17 group belonged to the sargramostim group.In patients given a 500 mg CCD, the median number of days to clinical improvement was significantly smaller in the high DCCL17 group (7.0 days) than in the low DCCL17 group (7.5 days) (Figure 6B, Figure S2, p ¼ 0.010).There were no significant differences in the total CCD ( 500 mg) that patients in the high and low DCCL17 groups received on Days 1-5 (206 ± 80 mg and 243 ± 115 mg, respectively; p ¼ 0.251).Conversely, in patients given a CCD >500 mg, the median number of days to clinical improvement was significantly smaller in the low DCCL17 group (7.0 days) than in the high DCCL17 group (11.0 days) (Figure 6C, p ¼ 0.017).There was a significant difference in the CCD (>500 mg) that the patients in the high and low DCCL17 groups received on Days 1-5 (1609 ± 456 mg and 2693 ± 1188 mg, respectively; p ¼ 0.008).As shown in Figure 6, in patients given a CCD 500 mg, the improvement in DOO was faster in the high DCCL17 group than in the low DCCL17 group (Figure 6D).However, the DOO in patients with a CCD >500 mg improved faster in the low DCCL17 group than in the high DCCL17 group (Figure 6E).Thus, sargramostim inhalation likely caused an increase in serum CCL17 when the CCD was 500 mg and might be associated with clinical improvement.The results of the confounding analysis are shown in Table S2.

Effect of sargramostim inhalation on HRCT characteristics
Pulmonary inflammation was evaluated by the % volume of opacities on HRCT according to the statistical analysis plan, but only 48 of the 70 patients were able to undergo imaging, mainly due to preventing nosocomial infections.Histograms of % volume corresponding to each Hounsfield unit (HU) bin ranging from minimum to maximum CT density were automatically extracted by SYNAPSE VINCENT.HRCT images and histograms of % volume, which was defined as the percentage of the lung volume of the HRCT image corresponding to a specific CT density range to the lung volume of the whole CT density range, corresponding to each HU bin on Day 0 (green line) and Day 5 (black line) in a typical patient with clinical improvement and marked reduction in consolidation on HRCT are shown in Figure 7A.The % volumes corresponding to ground-glass opacity (À800 to À550 HU) and consolidation (À550 to 200 HU) decreased towards Day 5 and shifted towards the normal lung CT density range of À1000 to À800 HU.
We also conducted a comparison of the histograms of D% volume, defined as the increase in % volume from Days 1-5.The proportion of À800 HU or more, corresponding to ground-glass opacity and consolidation due to lung inflammation, decreased in both groups from Days 1-5, and the proportion of À800 HU or less, corresponding to healthy lungs, tended to increase (Figure 7B).The rate of À800 HU or less seemed to be higher in the placebo group than in the sargramostim group, but the difference was not significant.
As a post hoc analysis, a comparison was made between the sargramostim and placebo groups in the 500 mg and >500 mg CCD groups (Figure 7C, D).In the CCD 500 mg group, D% volume tended to be  Improvement in the degree of oxygenation was delayed in patients who received a cumulative corticosteroid dose >500 mg.In cases of higher cumulative corticosteroid doses, there was an 'early dip' in which early oxygenation was impaired.Degree of oxygenation: 1 ¼ requiring supplemental oxygen delivered at !4 L/min using an oxygen mask or nasal high flow; 2 ¼ requiring supplemental oxygen at !2 L/min using a nasal cannula; 3 ¼ requiring supplemental oxygen at <2 L/min using a nasal cannula; 4 ¼ no supplemental oxygen required, but the patient reports shortness of breath when walking or moving; 5 ¼ no supplemental oxygen required and no shortness of breath when walking or moving.(B) Clinical improvement was suggested to be delayed with increasing cumulative corticosteroid dose over Days 1-5.Modified 7-category ordinal scale: 1 ¼ patient death; 2 ¼ use of invasive ventilation or extracorporeal membrane oxygenation; 3 ¼ use of non-invasive ventilation/high-flow oxygen supply device; 4 ¼ supplemental oxygen required; 5 ¼ no supplemental oxygen required, but continuous treatment required (for COVID-19 or other conditions); 6 ¼ without supplemental oxygen and the dose of corticosteroid was tapered to 25 mg prednisolone-equivalent; 7 ¼ patient discharged from hospital.Kaplan-Meier curves show the percentage of patients (vertical axis) who achieved improvement of two or more categories from baseline on a modified 7-category ordinal scale.In patients who received 500 mg cumulative corticosteroid dose (A), the median number of days to clinical improvement was smaller in the sargramostim group (6 days) than in the placebo group (7.5 days); this relationship was reversed in patients who received > 500 mg cumulative corticosteroid dose (B) (10.5 and 7 days for sargramostim and placebo groups, respectively).There was a significant difference (p ¼ 0.007) in the number of days for clinical improvement among the four groups; i.e., sargramostim or placebo with or without high-dose corticosteroids.Modified 7-category ordinal scale: 1 ¼ patient death; 2 ¼ use of invasive ventilation or extracorporeal membrane oxygenation; 3 ¼ use of non-invasive ventilation/high-flow oxygen supply device; 4 ¼ supplemental oxygen required; 5 ¼ no supplemental oxygen required, but continuous treatment required (for COVID-19 or other conditions); 6 ¼ without supplemental oxygen and the dose of corticosteroid was tapered to 25 mg prednisolone-equivalent; 7 ¼ patient discharged from hospital.Changes in the degree of oxygenation (vertical axis) are shown in patients who received 500 mg (C) or >500 mg (D) cumulative corticosteroid doses during Days 1-5 over days since enrolment (horizontal axis).Data are shown as mean; error bars indicate standard error.In patients who received 500 mg cumulative corticosteroid dose, improvement in the degree of oxygenation was faster in the sargramostim group than placebo group, whereas this relationship was reversed in patients who received >500 mg cumulative corticosteroid dose.Degree of oxygenation: 1 ¼ requiring supplemental oxygen delivered at !4 L/min using an oxygen mask or nasal high flow; 2 ¼ requiring supplemental oxygen at !2 L/min using a nasal cannula; 3 ¼ requiring supplemental oxygen at <2 L/min using a nasal cannula; 4 ¼ no supplemental oxygen required, but the patient reports shortness of breath when walking or moving; 5 ¼ no supplemental oxygen required and no shortness of breath when walking or moving.(A) The horizontal axis shows study days since enrolment.The vertical axis is the serum CCL17 level (pg/mL).Data are shown as mean (95% CI).Serum CCL17 level was significantly increased in the sargramostim group compared with the placebo group.Kaplan-Meier curves show the percentages of patients who achieved improvement of two or more categories from baseline on a modified 7-category ordinal scale stratified by DCCL17.Cumulative incidence in patients who received 500 mg and >500 mg cumulative corticosteroid doses are shown in (B) and (C), respectively.In patients who received a cumulative corticosteroid dose 500 mg, the clinical improvement was significantly faster in the high DCCL17 group than in the low DCCL17 group.This relationship was reversed in patients who received a cumulative corticosteroid dose >500 mg.Modified 7-category ordinal scale: 1 ¼ patient death; 2 ¼ use of invasive ventilation or extracorporeal higher in the sargramostim group than in the placebo group in the region of À900 HU or less.The D% volume in the region of À50 HU or higher was lower in the sargramostim group than in the placebo group in all CCD groups, but the difference was slightly larger in the 500 mg group than in the >500 mg group.According to the statistical analysis plan, the proportion of slices with the following findings in 18 slices of each HRCT was determined by visual scoring and compared between the sargramostim and placebo groups.The proportion of crazy-paving pattern, bronchial wall thickening, vascular thickening and airspace consolidation decreased from Days 0-5 in both groups (Figure 7E).Conversely, the proportion of subpleural curvilinear shadow and plate atelectasis increased.For all nine findings, the change was not significantly different between the sargramostim and placebo groups.
When a post hoc analysis was applied to the change in the proportion of each finding by corticosteroid dose from Days 1-5 (Figure 7F, G), the change of crazy-paving pattern was significantly decreased in the placebo vs sargramostim group in patients receiving a 500 mg CCD (mean ± standard error; À15.7%±9.2% and À2.6%±2.1%,respectively; p ¼ 0.048, Figure 7F).In patients with a CCD >500 mg, the proportion of plate atelectasis was significantly increased in the placebo vs sargramostim group (mean ± standard error; 12.7%±3.6%and 0.3%±1.6%,respectively; p ¼ 0.002, Figure 7G).Thus, the evaluation of pulmonary inflammation by HRCT appeared to be affected by the concomitant corticosteroid dose, combined with the time factor from Days 1-5, and whether the patient was in the sargramostim group or the placebo group.The results of the confounding analysis are shown in Table S3-A.
The effect of the total amount of changes in serum CCL17 during Days 1-5 on the D% volume was also analysed using histograms for the mean change in % volume corresponding to each 25-HU bin in patients who received sargramostim or placebo and received a CCD 500 mg over Days 1-5.For patients with DCCL17 < 1277 pg/mL in the sargramostim group, D% volume was higher in the CT range of À1000 to À800 HU and lower in both À800 to À550 HU and À200 to 100 HU than those with DCCL17 < 1277 pg/mL, suggesting that the % volume corresponding to ground-glass opacity and consolidation decreased and the % volume corresponding to normal lung tissue increased (Figure S3).Using a generalised estimation equation with D% volume as the objective variable, we confirmed that the interaction term between the treatment group, CT density, total steroid dose, and serum DCCL17 were all significant (Table S3-B, p < 0.001).

Safety outcomes
AEs occurred in 29 patients (61.7%) in the sargramostim group and in 11 patients (47.8%) in the placebo group.There were 47 events in the sargramostim group and 21 in the placebo group.AEs with an incidence >5% were constipation, increased white blood cell count, hyperglycaemia, and hepatocellular injury in the sargramostim group, and constipation and insomnia in the placebo group.There were two SAEs (myocardial infarction and COVID-19 pneumonia exacerbation) in the sargramostim group.The patient who had myocardial infarction recovered, and the patient with worsening COVID-19 pneumonia recovered on Day 55 but required oxygen on exertion.A causal relationship to the study drug was ruled out for both SAEs.

Discussion
This study was a placebo-controlled, randomised trial investigating the effects of short-term inhaled sargramostim on moderate COVID-19 pneumonia without Figure 6.(Continued) membrane oxygenation; 3 ¼ use of non-invasive ventilation/high-flow oxygen supply device; 4 ¼ supplemental oxygen required; 5 ¼ no supplemental oxygen supply required, but continuous treatment required (for COVID-19 or other conditions); 6 ¼ without supplemental oxygen and the dose of corticosteroid was tapered to 25 mg prednisolone-equivalent; 7 ¼ patient discharged from hospital.Changes in the degree of oxygenation are shown in patients who received 500 mg (D) or >500 mg (E) cumulative corticosteroid doses for Days 1-5, stratified by DCCL17.Data are shown as mean; error bars indicate standard error.In patients who received a cumulative corticosteroid dose 500 mg, a faster improvement in the degree of oxygenation was observed in the high DCCL17 group when compared with the low DCCL17 group, whereas this relationship was reversed in patients who received a cumulative corticosteroid dose >500 mg.Degree of oxygenation: 1 ¼ requiring supplemental oxygen delivered at !4 L/min using an oxygen mask or nasal high flow; 2 ¼ requiring supplemental oxygen at !2 L/min using a nasal cannula; 3 ¼ requiring supplemental oxygen at <2 L/min using a nasal cannula; 4 ¼ no supplemental oxygen required, but the patient reports shortness of breath when walking or moving; 5 ¼ no supplemental oxygen required and no shortness of breath when walking or moving.CCL: C-C motif chemokine ligand; CI: confidence interval.The rate of À800 HU or less seemed to be higher in the placebo group than in the standardising concomitant corticosteroid dose.According to the primary analysis, sargramostim inhalation neither hastened clinical improvement nor improved oxygenation.However, a post hoc sub-analysis found that sargramostim inhalation significantly accelerated clinical improvement and improved oxygenation compared with placebo inhalation when the CCD was 500 mg prednisolone-equivalent on Days 1-5.
It is important to consider the biological rationale behind the fact that the point estimates showed a numerically worse outcome for the sargramostim group than the placebo group, although there was no statistical difference between them.We think this is mainly because the treatment of discontinued cases was occasionally biased towards the sargramostim group, and thus, the sargramostim group appeared to be inferior to the placebo group in point estimates.
Around the time the protocol for this study was developed and the first patients were enrolled, the NIH guidelines had recommended corticosteroid doses for treatment of COVID-19 [17].However, there was no uniformity in the use of concomitant corticosteroids among pulmonary physicians in Japan.In fact, bolus methylprednisolone therapy was administered at three of the seven participating centres in this study.This clinical trial was conducted during the 3 rd to 5 th waves of the COVID-19 pandemic, which saw a rapid increase in deaths and a shortage of hospital beds.If the study headquarters had required the physicians of each facility to standardise the dose of concomitant corticosteroids, some facilities would have withdrawn from the study.CCD bias might not have occurred if patients had been assigned to the sargramostim and placebo groups by predicting the corticosteroid doses that they would receive on Days 1-5 and evenly assigning the doses.However, it was not realistic to predict CCD at registration by examining the ever-changing conditions of patients.As a result, the six patients who received a CCD of >2500 mg on Days 1-5 were biased towards the sargramostim group.This is why we conducted a post hoc analysis in this study.
The greatest enigma we discovered was the early oxygenation impairment that was more marked in patients receiving bolus methylprednisolone therapy.The mean DOO decreased during the first 5 days in patients who received a >2500 mg prednisolone-equivalent CCD on Days 1-5.Baseline DOO was not significantly different between groups with a CCD >2500 mg and those with a CCD 2500 mg.Neither the total prebaseline steroid dose nor the total number of treatment days were different between the two groups.The risk of viral reactivation has been reported in patients receiving high a CCD [18,19].Alternatively, bolus methylprednisolone therapy may prevent monocytes from differentiating into normal mature alveolar macrophages, which have been indicated to be essential for recovery from viral pneumonia [16].
According to the NIH COVID-19 treatment guidelines, the recommended corticosteroid dose for hospitalised patients with COVID-19 who require supplemental oxygen, including patients receiving non-invasive or mechanical ventilation, is 6 mg dexamethasone for 10 days [17], which is the equivalent of 400 mg prednisolone.The RECOVERY trial compared the outcome of low-dose dexamethasone (6 mg for 10 days) with high-dose dexamethasone (20 mg/day for 5 days followed by 10 mg/day for 5 days, which was 200 mg vs 500 mg prednisolone equivalent for 5 days) in patients with hypoxia receiving oxygen or SpO 2 <92% but did not require a ventilator [20], resulting in an increased risk of death in the higher-dose corticosteroid group.In contrast, 21 patients received a 500-2500 mg CCD from Days 1-5, and six patients received a >2500 mg CCD in this study.Because these far sargramostim group, but the difference was not significant.In the cumulative steroid dose 500 mg group, D% volume tended to be higher in the sargramostim group than in the placebo group in the region of À900 HU or less.D% volume in the region of À50 HU or higher was lower in the sargramostim group than in the placebo group in all cumulative steroid dose groups, but the difference was slightly larger in the 500 mg group than in the >500 mg group.Changes in the proportion of slices with crazy-paving pattern (CPA), bronchial wall thickening (BWT), subpleural curvilinear shadow (SCS), vascular thickening (VT), airspace consolidation (AC), small nodule opacity (SNO), reticular shadow (RS), plate atelectasis (PA), acute interstitial pneumonitis/acute respiratory distress syndrome pattern (AIP/ARDS) in 18 slices of each HRCT from Days 0-5 are shown in all 48 patients (E), patients who received 500 mg (F) and those who received >500 mg (G) cumulative corticosteroid doses for Days 1-5.For all nine findings, the change in proportion of axial slices with each finding was not significantly different between the sargramostim and placebo groups.When we evaluated the change in the proportion of each finding by corticosteroid dose from Days 1-5, the proportion of axial slices with CPA was significantly decreased in the placebo vs sargramostim group in patients receiving a 500 mg cumulative corticosteroid dose.In patients with a cumulative corticosteroid dose >500 mg, the proportion of axial slices with PA was significantly increased in the placebo group vs the sargramostim group.HRCT: high-resolution computed tomography; HU: Hounsfield unit exceed the corticosteroid doses used in the RECOVERY trial, it was difficult to simply compare the effects of corticosteroids on COVID-19 pneumonia.However, considering that >2500 mg corticosteroid administration during 5 days caused early exacerbation in oxygenation in this study, high-dose corticosteroid administration for patients with moderate COVID-19 pneumonia might be a treatment that should be avoided.
Most reports, except for case studies, have used visual scoring by radiologists to evaluate HRCT images for COVID-19 pneumonia [21,22].Visual scoring facilitates the evaluation of various changes and lesions that could not have been distinguished using artificial intelligence (AI) analysis.Although a few reports using AI analysis had the advantage of being able to quantitatively evaluate the range and degree of inflammation in the whole lung [23,24], several outstanding issues preclude the routine use of AI analysis in clinical practice.For example, 10 axial slices, originating from 7 of the 58 cases of HRCT-DICOM data, which were analysed in the present study, required manual visual correction of the lung fields.The errors requiring manual intervention were caused by the failure of the automatic lung field extraction function to recognise parts of the emphysematous region or consolidations as comprising the lung field.Nevertheless, unless there was a large loss or excess, these errors would not have affected the results of the analysis.
GM-CSF accelerates the differentiation of recruited monocytes into mature alveolar macrophages and is important for maintaining lung homeostasis [16].The results from single-cell RNA sequencing of alveolar macrophages from bronchoalveolar lavage in patients with COVID-19 pneumonia revealed that GM-CSF-mediated differentiation into mature alveolar macrophages was insufficient in these patients.This resulted in an increase in alternatively differentiated macrophages, triggering a cytokine storm.Other researchers recognised GM-CSF to be a crucial factor of CD8 þ T lymphocytes to activate dendritic cells during priming in lymphoid tissue, providing a positive feedback loop in the stimulation of CD8 þ T cell proliferation in viral infection [25].In this regard, GM-CSF administration initiates lung CD8 þ T cell responses by maintaining tissue-resident CD103 þ dendritic cells [26].Notably, mice administered anti-GM-CSF antibodies are reported to be more susceptible to SARS-CoV-2 infection and have reduced survival [27].Moreover, GM-CSF has been reported to have a life-saving effect in lethal influenza A-infected mice by improving host defence function and moderating M1-like airway monocyte/macrophage polarisation [13,28].
CCL17 is a chemokine that is predominantly expressed in airway macrophages in response to the overexpression of GM-CSF in lungs infected with the influenza A virus.Moreover, CCL17 is known to be a prognostic factor in patients with COVID-19; i.e. higher levels of serum CCL17 are associated with better prognosis [29].In the present study, among patients who received a 500 mg CCD during Days 1-5, clinical improvement was significantly faster in those with higher serum DCCL17.Because the group with higher serum DCCL17 coincided with the sargramostim inhalation group, it appears that inhaled sargramostim elicited an increase in serum DCCL17.However, this relationship changed when the corticosteroid dose was increased.In patients receiving a >500 mg CCD, clinical improvement was independent of sargramostim inhalation; serum CCL17 and clinical improvement were inversely correlated with the CCD.This phenomenon was confirmed by a generalised estimating equation of changes in the treatment group, CCD, and CCL17, with respect to the change in DOO or modified 7-category ordinal scale.The results suggest that high-dose corticosteroid administration inhibits the release of CCL17 and impairs the differentiation of monocytes into mature macrophages.
In conclusion, the present study suggests that sargramostim inhalation might improve oxygenation and hasten clinical recovery in patients with moderate II COVID-19.However, this effect was limited to cases where concomitantly administered corticosteroids were within NIH-recommended doses.We believe that this report will contribute to the successful treatment of moderate II COVID-19 pneumonia and the prevention of its progression to ARDS.access to all the data in the study and accept responsibility to submit the manuscript for publication.

Figure 1 .
Figure 1.Study flow.Of the 76 patients with COVID-19 pneumonia who underwent screening, 75 were enrolled in the study.Three patients in the sargramostim group and two patients in the placebo group withdrew from the trial prior to administration of the study drug, and 47 and 23 patients received treatment with sargramostim or placebo, respectively.Three patients in the sargramostim group withdrew from the study, and 44 and 23 patients completed the study by Day 28 in the sargramostim and placebo groups, respectively.Sargramostim: saccharomycesderived recombinant granulocyte-macrophage colony-stimulating factor, ICU: intensive care unit.

Figure 2 .
Figure 2. Clinical improvement on an ordinal scale.Kaplan-Meier curves showing the percentage of patients (vertical axis) who achieved improvement in two or more categories on the modified 7-category ordinal scale from baseline over days since enrolment (horizontal axis) in the sargramostim (blue line) and placebo (red line) groups.There was no difference between the sargramostim and placebo groups.Modified 7-category ordinal scale: 1 ¼ patient death; 2 ¼ use of invasive ventilation or extracorporeal membrane oxygenation; 3 ¼ use of non-invasive ventilation/high-flow oxygen supply device; 4 ¼ supplemental oxygen required; 5 ¼ no supplemental oxygen required, but continuous treatment required (for COVID-19 or other conditions); 6 ¼ without supplemental oxygen and the dose of corticosteroid was tapered to 25 mg prednisolone-equivalent; and 7 ¼ patient discharged from hospital.

Figure 3 .
Figure 3. Degree of oxygenation (vertical axis) in patients over days since enrolment (horizontal axis) in sargramostim (blue line) and placebo (red line) groups.Data are shown as mean; error bars indicate standard deviation.The degree of oxygenation in the sargramostim group did not differ from the placebo group.Degree of oxygenation: 1 ¼ requiring supplemental oxygen delivered at !4 L/min using an oxygen mask or nasal high flow; 2 ¼ requiring supplemental oxygen at !2 L/min using a nasal cannula; 3 ¼ requiring supplemental oxygen at <2 L/min using a nasal cannula; 4 ¼ no supplemental oxygen required, but the patient reports shortness of breath when walking or moving; 5 ¼ no supplemental oxygen required and no shortness of breath when walking or moving.

Figure 4 .
Figure 4. Time courses of the degree of oxygenation (A), and Kaplan-Meier curves showing the percentage of patients who achieved improvement of two or more categories on a modified 7-category ordinal scale from baseline (B), stratified by cumulative corticosteroid dose ( 500 mg: blue line, >500-2500 mg: orange line, and >2500 mg: green line, for Days 1-5) over days since enrolment.(A) Data are shown as mean; error bars indicate standard error.Improvement in the degree of oxygenation was delayed in patients who received a cumulative corticosteroid dose >500 mg.In cases of higher cumulative corticosteroid doses, there was an 'early dip' in which early oxygenation was impaired.Degree of oxygenation: 1 ¼ requiring supplemental oxygen delivered at !4 L/min using an oxygen mask or nasal high flow; 2 ¼ requiring supplemental oxygen at !2 L/min using a nasal cannula; 3 ¼ requiring supplemental oxygen at <2 L/min using a nasal cannula; 4 ¼ no supplemental oxygen required, but the patient reports shortness of breath when walking or moving; 5 ¼ no supplemental oxygen required and no shortness of breath when walking or moving.(B) Clinical improvement was suggested to be delayed with increasing cumulative corticosteroid dose over Days 1-5.Modified 7-category ordinal scale: 1 ¼ patient death; 2 ¼ use of invasive ventilation or extracorporeal membrane oxygenation; 3 ¼ use of non-invasive ventilation/high-flow oxygen supply device; 4 ¼ supplemental oxygen required; 5 ¼ no supplemental oxygen required, but continuous treatment required (for COVID-19 or other conditions); 6 ¼ without supplemental oxygen and the dose of corticosteroid was tapered to 25 mg prednisolone-equivalent; 7 ¼ patient discharged from hospital.

Figure 5 .
Figure 5. Clinical improvement on ordinal scale (A, B) and changes in degree of oxygenation (C, D), stratified by cumulative corticosteroid dose ( 500 mg for the left or >500 mg for the right during days 1-5) over days since enrolment.Kaplan-Meier curves show the percentage of patients (vertical axis) who achieved improvement of two or more categories from baseline on a modified 7-category ordinal scale.In patients who received 500 mg cumulative corticosteroid dose (A), the median number of days to clinical improvement was smaller in the sargramostim group (6 days) than in the placebo group (7.5 days); this relationship was reversed in patients who received > 500 mg cumulative corticosteroid dose (B) (10.5 and 7 days for sargramostim and placebo groups, respectively).There was a significant difference (p ¼ 0.007) in the number of days for clinical improvement among the four groups; i.e., sargramostim or placebo with or without high-dose corticosteroids.Modified 7-category ordinal scale: 1 ¼ patient death; 2 ¼ use of invasive ventilation or extracorporeal membrane oxygenation; 3 ¼ use of non-invasive ventilation/high-flow oxygen supply device; 4 ¼ supplemental oxygen required; 5 ¼ no supplemental oxygen required, but continuous treatment required (for COVID-19 or other conditions); 6 ¼ without supplemental oxygen and the dose of corticosteroid was tapered to 25 mg prednisolone-equivalent; 7 ¼ patient discharged from hospital.Changes in the degree of oxygenation (vertical axis) are shown in patients who received 500 mg (C) or >500 mg (D) cumulative corticosteroid doses during Days 1-5 over days since enrolment (horizontal axis).Data are shown as mean; error bars indicate standard error.In patients who received 500 mg cumulative corticosteroid dose, improvement in the degree of oxygenation was faster in the sargramostim group than placebo group, whereas this relationship was reversed in patients who received >500 mg cumulative corticosteroid dose.Degree of oxygenation: 1 ¼ requiring supplemental oxygen delivered at !4 L/min using an oxygen mask or nasal high flow; 2 ¼ requiring supplemental oxygen at !2 L/min using a nasal cannula; 3 ¼ requiring supplemental oxygen at <2 L/min using a nasal cannula; 4 ¼ no supplemental oxygen required, but the patient reports shortness of breath when walking or moving; 5 ¼ no supplemental oxygen required and no shortness of breath when walking or moving.

Figure 6 .
Figure 6.Change in mean serum CCL17 levels (A) from Days 1-5 in patients receiving inhaled sargramostim (n 5 47) and placebo (n 5 22), and clinical improvement on an ordinal scale (B, C) and changes in degree of oxygenation (D, E), stratified by above or below the median change (1277 pg/mL) of serum CCL17 level (DCCL17).(A)The horizontal axis shows study days since enrolment.The vertical axis is the serum CCL17 level (pg/mL).Data are shown as mean (95% CI).Serum CCL17 level was significantly increased in the sargramostim group compared with the placebo group.Kaplan-Meier curves show the percentages of patients who achieved improvement of two or more categories from baseline on a modified 7-category ordinal scale stratified by DCCL17.Cumulative incidence in patients who received 500 mg and >500 mg cumulative corticosteroid doses are shown in (B) and (C), respectively.In patients who received a cumulative corticosteroid dose 500 mg, the clinical improvement was significantly faster in the high DCCL17 group than in the low DCCL17 group.This relationship was reversed in patients who received a cumulative corticosteroid dose >500 mg.Modified 7-category ordinal scale: 1 ¼ patient death; 2 ¼ use of invasive ventilation or extracorporeal

Figure 7 .
Figure 7. Change in high-resolution computed tomography characteristics for Days 0-5.(A) HRCT images and histograms of % volume corresponding to each HU bin on Day 0 (green line) and Day 5 (black line) in a typical patient with clinical improvement.The % volumes corresponding to ground-glass opacity (À800 to À550 HU) and consolidation (À550 to 200 HU) decreased during Days 0-5 and shifted towards CT density range of À1000 to À800 HU.The percent volume was defined as the percentage of the lung volume of the HRCT image corresponding to a specific CT density range to the lung volume of the whole CT density range (minimum to maximum CT value automatically extracted by SYNAPSE VINCENT).Mean changes in % volume corresponding to each 25-HU bin are shown for Days 0À5 in all 48 patients (B), those who received 500 mg (C) and those who received >500 mg (D) cumulative corticosteroid doses for Days 1-5.The rate of À800 HU or less seemed to be higher in the placebo group than in the

Figure 7 .
Figure 7. (Continued) sargramostim group, but the difference was not significant.In the cumulative steroid dose 500 mg group, D% volume tended to be higher in the sargramostim group than in the placebo group in the region of À900 HU or less.D% volume in the region of À50 HU or higher was lower in the sargramostim group than in the placebo group in all cumulative steroid dose groups, but the difference was slightly larger in the 500 mg group than in the >500 mg group.Changes in the proportion of slices with crazy-paving pattern (CPA), bronchial wall thickening (BWT), subpleural curvilinear shadow (SCS), vascular thickening (VT), airspace consolidation (AC), small nodule opacity (SNO), reticular shadow (RS), plate atelectasis (PA), acute interstitial pneumonitis/acute respiratory distress syndrome pattern (AIP/ARDS) in 18 slices of each HRCT from Days 0-5 are shown in all 48 patients (E), patients who received 500 mg (F) and those who received >500 mg (G) cumulative corticosteroid doses for Days 1-5.For all nine findings, the change in proportion of axial slices with each finding was not significantly different between the sargramostim and placebo groups.When we evaluated the change in the proportion of each finding by corticosteroid dose from Days 1-5, the proportion of axial slices with CPA was significantly decreased in the placebo vs sargramostim group in patients receiving a 500 mg cumulative corticosteroid dose.In patients with a cumulative corticosteroid dose >500 mg, the proportion of axial slices with PA was significantly increased in the placebo group vs the sargramostim group.HRCT: high-resolution computed tomography; HU: Hounsfield unit

Table 1 .
Patient baseline demographics and clinical characteristics.
a Student's t-test for continuous variables; chi-square test for categorical variables.