Rebound Hypoxemia in Mechanically Ventilated Patients With COVID-19 Completing a Standard 10-Day Course of Corticosteroid Therapy

Abstract

Background:

Rebound hypoxemia may occur after cessation of corticosteroid therapy for COVID-19 pneumonitis. We aimed to determine the incidence of this phenomenon in mechanically ventilated patients with COVID-19 completing corticosteroid therapy.

Methods:

We conducted a retrospective observational cohort study across 2 tertiary ICUs from September to December 2021. We included all adult patients receiving invasive mechanical ventilation on completion of a 10-day course of dexamethasone for COVID pneumonitis. Our primary outcome was change in P_aO2/F_IO2 ratio in the 7 days following cessation of dexamethasone. Secondary outcomes included duration of ventilation, frequency of corticosteroid recommencement, and mortality.

Results:

We studied 88 subjects. Median age was 61 years (interquartile range [IQR] 51–67), and median duration of mechanical ventilation was 14 days (IQR 11–26). On completion of dexamethasone, 62 subjects (70%) remained off corticosteroid therapy for the subsequent 7 days. Of these, 63% (n = 20) had a stable P_aO2/F_IO2 ratio on day 7 (mean change 197 ± 42 mm Hg to 258 ± 83 mm Hg). Subjects who had a reduction in P_aO2/F_IO2 ratio >20 mm Hg by day 7 (mean change 214 ± 68 mm Hg to 152 ± 78 mm Hg) were more likely to die in hospital (P < .001). These subjects had a higher C-reactive protein level at time of steroid cessation (204 ± 87 mg/L, P = .17), which remained persistently elevated (206 ± 89 mg/L, P = .01) on day 7. The 30% of subjects who recommenced steroids required a longer duration of ventilation (13.5 vs 24.5 d, P = .002) but showed no differences in ICU mortality (19% vs 27%, P = .43).

Conclusions:

Rebound hypoxemia and steroid recommencement were common occurrences in mechanically ventilated subjects with COVID-19. Rebound hypoxemia was associated with higher mortality. Steroid recommencement was associated with longer duration of mechanical ventilation but no significant difference in mortality.

Introduction

Current guidelines support the use of corticosteroid therapy in hospitalized patients with SARS-CoV-2 infection.^1,2 The landmark RECOVERY trial showed the use of dexamethasone for up to 10 days for subjects receiving supplemental oxygen therapy or invasive mechanical ventilation resulted in lower 28-day mortality.³ However, optimal duration of corticosteroid therapy in this patient cohort remains unclear.⁴

Observational studies have described the occurrence of rebound hypoxemia in a subgroup of patients after steroid cessation,^5,6 as has been previously described in non-COVID-19 patients with ARDS.⁷ Furthermore, there are case reports describing steroid-responsive organizing pneumonia in patients who fail to improve or deteriorate after steroid cessation.^8–10

Overall, there are limited data describing the characteristics and clinical outcomes of steroid therapy responders who have rebound respiratory deterioration. In particular, there are limited data available for critically unwell patients who are unable to be liberated from mechanical ventilation after 10 days of corticosteroid treatment.

We aimed to determine the effect of a 10-day course of dexamethasone therapy and incidence of rebound hypoxemia in mechanically ventilated patients with COVID-19. We hypothesized that a proportion of patients with COVID-19 who completed a 10-day course of dexamethasone therapy and remained invasively ventilated would develop decreasing P_aO2/F_IO2 in the 48-hour to 7-day period following cessation of dexamethasone in comparison to the last 48 hours on dexamethasone therapy.

QUICK LOOK

Current knowledge

Corticosteroid use in critically ill patients with COVID-19 is well established and associated with mortality benefit. Current international guidelines suggest the use of intravenous or oral corticosteroids for 7–10 days for this population.

What this paper contributes to our knowledge

In our study of critically ill subjects with COVID-19, rebound hypoxemia following steroid cessation was relatively common and associated with higher ICU mortality. These subjects had a persistently elevated C-reactive protein at day 7 following steroid cessation. Daily C-reactive protein measurement may assist clinicians in monitoring the trajectory of systemic inflammation following cessation of corticosteroids, and a persistently raised C-reactive protein may justify consideration of recommencing therapy. Recommencing steroids was not associated with higher mortality in our study.

Methods

Study design and setting

We conducted a retrospective observational cohort study across 2 adult tertiary ICUs in Melbourne, Australia. We analyzed 4 months of patient admissions from September 1 to December 31, 2021, and included all adult ICU patients (age ≥18 years) with confirmed SARS-CoV-2 infection by reverse transcription polymerase chain reaction testing who were receiving invasive mechanical ventilation on completion of a standard 10-day course of dexamethasone for COVID pneumonitis.

Data collection

Data were collected via electronic medical record audit performed by members of the research team. This included basic demographic data, vaccination status, hospital and ICU admission and discharge details, intensive care treatments and use of corticosteroid, antimicrobial and immunomodulatory therapies, treatment complications, physiological and biochemical parameters, mechanical ventilation settings, and computed tomography (CT) reports.

We obtained daily data on specific respiratory parameters from 48 hours prior to cessation of dexamethasone up to 7 days post-cessation of corticosteroid therapy. This included arterial blood gases with the lowest P_aO2/F_IO2 in each 24-hour period as well as ventilation mode, peak inspiratory pressure (PIP), PEEP, breathing frequency, and tidal volume at 0800 each day.

Outcomes

We assessed oxygenation outcomes by comparing changes in the P_aO2/F_IO2 in the 7 days following cessation of dexamethasone to subject baseline, defined as the final day of a standard dexamethasone course. We defined “rebound hypoxemia” as a reduction in P_aO2/F_IO2 by >20 mm Hg from baseline measurement¹¹ on any day following cessation of dexamethasone. Changes in respiratory parameters such as PIP, PEEP, and breathing frequency were also analyzed.

Secondary outcomes included duration of mechanical ventilation, length of ICU stay, and ICU and hospital mortality. We also examined the use of alternative corticosteroids, frequency of corticosteroid recommencement, and the incidence of infective complications and organizing pneumonia.

Statistical analysis

Data were analyzed using STATA version 14.2 (Stata, College Station, Texas). Continuous variables are reported as mean (standard deviation) if normally distributed or median (interquartile range [IQR]) if non-normally distributed. Continuous variables were compared using analysis of variance (if normally distributed) or the Wilcoxon rank-sum test (if non-normally distributed). Categorical variables are reported as frequency (percentages) and were compared using Pearson’s chi-square test or Fisher exact test as appropriate. Changes in daily variables over time were assessed using the Friedman test and presented as box plots. A mixed-effect model was developed to investigate factors associated with rebound hypoxemia on any individual day, with subject repeated measurements treated as a random effect. All variables were considered for inclusion in the multivariate analysis, with the final model including those with P values of <.1 on univariate analysis. Throughout all analyses, a P value <.05 was considered statistically significant. As data were generated through the process of routine care, the decision to complete investigations, such as CT and C-reactive protein measurement, was determined by the treating clinician and was not present for all subjects on all study days. Similarly, data regarding invasive ventilation were available only until subjects were liberated from mechanical ventilation. The number of subjects with corresponding data available is reflected in the denominator of relevant results. For mixed-effect modeling, complete-case analysis was performed, without imputation of missing data.

Ethics

Approval for the study with waiver of consent was granted (Project Reference No: HREC/84262/Austin-2022) by the Austin Health Human Research Ethics Committee, for both sites according to state regulations, on June 28, 2022.

Results

Subject characteristics

Over the 4-month study period, 88 critically ill subjects with COVID pneumonia remained on mechanical ventilation after completing a standard 10-day course of dexamethasone. The median age of the cohort was 61 years. Only 10% of the cohort had received 2 or more vaccine doses. Chronic respiratory comorbidities were infrequent. One third of subjects received antiviral therapy or Janus kinase inhibitor therapy. Use of interleukin 6 inhibitor and monoclonal antibody therapies was uncommon. The median duration of mechanical ventilation was 14 days. Less than one quarter of subjects died in hospital (Table 1). There were minimal missing data.

Table 1.

Subject characteristics

Characteristic	All subjects (N = 88)	Steroids ceased (n = 62)	Steroids ceased and PaO2/FIO2 Stable on day 7 (n = 20)	Steroids ceased and PaO2/FIO2 reduced on day 7 (n = 12)	P value*
Baseline characteristics
Age, years	61 (51–67)	60 (51–67)	59 (51–66)	65 (61–70)	.05
Male	53 (60)	38 (61)	10 (50)	9 (75)	.16
Height, cm	165 (158–175)	169 (159–176)	169 (157–176)	170 (155–177)	.91
Weight, kg	87 (76–106)	93 (78–108)	88 (75–111)	82 (65–94)	.26
BMI, kg/m2	32 (28–39)	33 (29–39)	34 (29–39)	29 (25–33)	.07
Comorbidities
Obesity	54 (61)	40 (65)	12 (60)	5 (42)	.31
Diabetes	33 (38)	24 (39)	7 (35)	2 (17)	.26
Receiving ACEi or ARB	24 (27)	18 (29)	5 (25)	3 (25)	>.99
Cardiac disease	10 (11)	9 (15)	3 (15)	2 (17)	>.99
Asthma	9 (10)	7 (11)	3 (15)	1 (8)	>.99
Pulmonary disease	7 (8)	4 (6)	0 (0)	3 (25)	.04
Chronic kidney disease	7 (8)	3 (5)	0 (0)	0 (0)	N/A
Immunosuppressed	5 (6)	2 (3)	0 (0)	1 (8)	.38
Rheumatological disease	4 (5)	2 (3)	0 (0)	2 (17)	.13
Vaccines received (no.)					.26
0	61 (69)	47 (76)	16 (80)	7 (58)
1	18 (21)	12 (19)	4 (20)	4 (33)
2	9 (10)	3 (5)	0 (0)	1 (8)
Treatments
Monoclonal antibody	5 (6)	3 (5)	2 (10)	1 (8)	>.99
Antiviral	32 (36)	26 (42)	7 (35)	7 (58)	.20
IL-6 inhibitor	7 (8)	5 (8)	4 (20)	0 (0)	.27
JAK inhibitor	31 (35)	23 (37)	6 (30)	8 (67)	.04
Outcomes
Duration of mechanical ventilation, days	14 (11–26)	14 (11–17)	15 (13–26)	18 (15–25)	.15
ICU stay, days	17 (13–27)	16 (13–21)	19 (17–26)	24 (18–28)	.23
ICU mortality	19 (22)	12 (19)	0 (0)	6 (50)	<.001
Hospital stay, days	25.5 (17.5–40)	25 (17–33)	32 (25–37)	31 (23–44)	.83
Discharge destination					.005
Died in hospital	20 (23)	12 (19)	0 (0)	6 (50)
Home	39 (44)	32 (52)	12 (60)	4 (33)
Other hospital	9 (10)	7 (11)	2 (10)	0 (0)
Rehabilitation	20 (23)	11 (18)	6 (30)	2 (17)

Data are shown as n (%) or median (interquartile range).

^* P values refer to comparison of subjects with stable and reduced P_aO2/F_IO2.

ACEi, angiotensin converting enzyme inhibitor; ARB, angiotensin II receptor blocker; BMI, body mass index; IL-6, interleukin 6; JAK, janus kinase.

More than two fifths of the cohort had a formal diagnosis of organizing pneumonia, based on CT scans, during their ICU stay. The most frequent radiological changes noted on initial CT were bilateral distribution of consolidation and ground glass opacity. A quarter of the subjects also had evidence of pulmonary thromboembolism (Table 2). The initial CT was performed at a median of 9 days (IQR 6–14) following commencement of dexamethasone. A follow-up CT was performed for almost two fifths of the cohort. This was performed at a median 21 days (IQR 15–26) (Supplementary Table S7). There were no statistically significant differences between groups on initial or subsequent CT scans for radiographic infective changes.

Table 2.

Disease characteristics

Characteristic	All subjects (N = 88)	Steroids ceased (n = 62)	Steroids ceased and PaO2/FIO2 stable on day 7 (n = 20)	Steroids ceased and PaO2/FIO2 reduced on day 7 (n = 12)	P value*
Formal diagnosis of organizing pneumonia, n (%)	39 (44)	27 (44)	13 (65)	7 (58)	.71
Formal diagnosis of pulmonary embolism, n (%)	22 (25)	17 (27)	10 (50)	5 (42)	.65
Microbiology, n (%)
Serum galactomannan	2 (2)	2 (3)	1 (5)	1 (8)	>.99
Bronchoalveolar lavage galactomannan	4 (5)	1 (2)	1 (5)	0 (0)	>.99
Initial CT chest findings, n (%)a	67 (76)	45 (73)	18 (90)	10 (83)	.58
Days since starting dexamethasone when CT performed (days), mean ± SD	10.5 ± 6.3	10.6 ± 6.5	12.6 ± 6.9	11.2 ± 7.8	.64
Bilateral distribution	63 (94)	43 (96)	17 (94)	10 (100)	.45
Peripheral or subpleural distribution	34 (51)	23 (51)	9 (50)	6 (60)	.61
Consolidation	56 (84)	35 (78)	15 (83)	10 (100)	.17
Traction bronchiectasis	6 (9)	3 (7)	2 (11)	1 (10)	>.99
Ground glass opacity	58 (87)	39 (87)	16 (89)	8 (80)	.52
Intralobular or interlobular septal thickening	7 (11)	5 (11)	3 (17)	1 (10)	>.99
Fibrosis or honeycombing	3 (5)	2 (4)	1 (6)	1 (10)	.66
Pulmonary embolism	14 (21)	11 (24)	6 (33)	3 (30)	.86
Organising pneumonia	14 (21)	10 (22)	6 (33)	3 (30)	.86

^* P values refer to comparison of subjects with stable and reduced P_aO2/F_IO2.

^a CT findings reported as percentage of subjects who underwent CT.

CT, computed tomography; SD, standard deviation.

Steroid cessation and incidence of rebound hypoxemia

On completion of a standard 10-day course of dexamethasone, 62 subjects (70%) remained off any corticosteroid therapy for the subsequent 7 days, whereas 26 subjects (30%) had an alternative corticosteroid commenced by treating clinicians within that time frame. Of those that remained off steroids, 32 subjects remained in the ICU for 7 days after dexamethasone cessation (Fig. 1). Almost two thirds of these subjects had a stable P_aO2/F_IO2 on day 7, defined as a P_aO2/F_IO2 within 20 mm Hg of their baseline measurement (mean change 197 ± 42 mm Hg to 258 ± 83 mm Hg) (Table 3).

Fig. 1.

Flow chart.

Table 3.

Daily characteristics on final day of steroid treatment

	Steroids ceased (n = 62)	Steroids ceased and PaO2/FIO2 stable on day 7 (n = 20)	Steroids ceased and PaO2/FIO2 reduced on day 7 (n = 12)	P value*
Respiratory variables
PEEP (cm H2O), Mean ± SD	10.6 ± 2.5	10.8 ± 2.6	11.8 ± 1.6	.20
FIO2, median (IQR)	0.35 (0.30–0.45)	0.35 (0.30–0.43)	0.38 (0.29–0.45)	.65
PaO2 (mm Hg), median (IQR)	72 (67–80)	71 (65–81)	71 (67–82)	.56
PaO2/FIO2, mean ± SD	207 ± 55	197 ± 42	214 ± 68	.38
PaCO2 (mm Hg), median (IQR)	43 (40–47)	43 (39–47)	43 (41–49)	.97
Breathing frequency (breaths/min), mean ± SD	20 ± 5	20 ± 4	18 ± 4	.18
Tidal volume (mL), mean ± SD	452 ± 104	440 ± 113	437 ± 110	.95
ΔP (cm H2O), mean ± SD	12.3 ± 3.3	13.1 ± 2.4	12.5 ± 3.5	.55
Prone, n (%)	4 (7)	2 (10)	0 (0)	.54
Paralysis, n (%)	13 (21)	5 (25)	3 (25)	>.99
Nitric oxide, n (%)	1 (2)	1 (5)	0 (0)	>.99
Ventilation mode, n (%)				.34
Continuous spontaneous ventilation	31 (50)	9 (45)	3 (25)
PC-IMV	23 (37)	8 (40)	8 (67)
VC-IMV	8 (13)	3 (15)	1 (8)
Other variables
Dexamethasone dose (mg), median (IQR)	6 (6–6)	6 (6–6)	6 (6–6)	N/A
Vasopressors, n (%)	19 (31)	5 (25)	6 (50)	.15
Antibiotics, n (%)	50 (81)	15 (75)	11 (92)	.24
Antifungals, n (%)	9 (15)	3 (15)	1 (8)	.58
C-reactive protein (mg/L), mean ± SD	105 ± 87	125 ± 80	204 ± 87	.17
WCC (×109/L), mean ± SD	11.7 ± 4.8	10.6 ± 3.0	11.3 ± 4.4	.57
Positive blood culture, n (%)	3 (5)	1 (5)	1 (8)	>.99

^* P values refer to comparison of patients with stable and reduced P_aO2/F_IO2.

IMV, invasive mechanical ventilation; IQR, interquartile range; PC-IMV, pressure control intermittent mandatory ventilation; VC-IMV, volume control intermittent mandatory ventilation; WCC, white cell count.

Subjects who had a reduction in P_aO2/F_IO2 >20 mm Hg by day 7 after cessation of steroids (mean change 214 ± 68 mm Hg to 152 ± 78 mm Hg) were older, had higher rates of chronic pulmonary disease, were more likely to have received Janus kinase inhibitors, and were more likely to die in the hospital (P < .001). These subjects had no differences in their ventilatory parameters, pneumonia characteristics, or adjunctive treatments administered on their final day of dexamethasone therapy that may have indicated increased risk of rebound hypoxemia. However, they had a higher C-reactive protein level at this time (mean 204 ± 87 mg/L, P = .17), which remained persistently elevated (mean 206 ± 89 mg/L, P = .01) on day 7 following cessation of steroids (Table 4, Fig. 2). A high proportion of subjects in both groups were receiving antibiotics at time of steroid cessation (81%) and remained on therapy at day 7 after steroid cessation (65%). There were no significant differences in use of antimicrobial therapy, number of positive blood cultures, or white cell count between the two groups.

Table 4.

Characteristics on day 7 after cessation of steroid treatment

	Steroids ceased (n = 32)	PaO2/FIO2stable on day 7 (n = 20)	PaO2/FIO2reduced on day 7 (n = 12)	P value*
Respiratory variables
PEEP (cm H2O)	9.9 ± 2.1	9.6 ± 2.0	10.2 ± 2.3	.51
FIO2	0.35 (0.30–0.42)	0.30 (0.30–0.38)	0.46 (0.40–0.68)	.01
PaO2 (mm Hg)	71 (64–89)	79 (67–99)	65 (59–72)	.02
PaO2/FIO2	218 ± 96	258 ± 83	152 ± 78	.001
PaCO2 (mm Hg)	39 (34–48)	38 (34–43)	47 (37–52)	.27
Frequency (breaths/min)	26 ± 8	25 ± 8	27 ± 7	.59
Tidal volume (mL)	378 ± 136	385 ± 93	369 ± 185	.79
ΔP (cm H2O)	13.9 ± 3.8	12.2 ± 1.8	16.3 ± 4.7	.01
Prone	0 (0)	0 (0)	0 (0)	N/A
Paralysis	5 (14)	1 (5)	4 (33)	.03
Nitric oxide	1 (3)	0 (0)	1 (8)	.38
Ventilation mode				.18
Noninvasive ventilation	17 (46)	9 (45)	3 (25)
Continuous spontaneous ventilation	13 (35)	9 (45)	4 (33)
PC-IMV	6 (16)	2 (10)	4 (33)
VC-IMV	1 (3)	0 (0)	1 (8)
Other variables
Dexamethasone dose (mg)	0 (0–0)	0 (0–0)	0 (0–0)	N/A
Vasopressors	13 (35)	5 (25)	8 (67)	.02
Antibiotics	24 (65)	13 (65)	10 (83)	.26
Antifungal	6 (16)	4 (20)	2 (17)	.82
C-reactive protein (mg/L)	150 ± 96	93 ± 62	206 ± 89	.01
WCC (×109/L)	10 ± 4	8 ± 3	11 ± 5	.07
Positive blood culture	1 (3)	1 (5)	0 (0)	>.99

Data are shown as mean ± SD, median (interquartile range), or n (%).

^* P values refer to comparison of subjects with stable and reduced P_aO2/F_IO2.

IQR, interquartile range; PC-IMV, pressure control intermittent mandatory ventilation; SD, standard deviation; VC-IMV, volume control intermittent mandatory ventilation; WCC, white cell count.

Fig. 2.

Daily C-reactive protein values for subjects who remained in ICU for 7 days after cessation of steroid treatment.

Steroid recommencement and subject outcomes

Subjects who were recommenced on steroids after completion of their dexamethasone therapy were similar at baseline and had no differences in their pneumonia characteristics to those who remained off steroids (Supplementary Table S3, Supplementary Table S4). On their final day of dexamethasone treatment, however, they had higher breathing frequencies, were receiving higher PEEP levels, and were more likely to be in a mandatory ventilation mode and to be receiving prone positioning (Supplementary Table S5).

The majority of subjects who were recommenced on steroids began receiving this therapy within 2 days of completing their course of dexamethasone. At the time of steroid recommencement, the mean P_aO2/F_IO2 was 175 mm Hg, mean PEEP 11 cm H₂O, and the median P_aCO2 was 48 mm Hg. Three quarters of these subjects were receiving mandatory ventilation, half were receiving neuromuscular paralysis, and one third were receiving prone positioning (Supplementary Table S6). There were no differences in rates of organizing pneumonia between groups.

Subjects who were recommenced on steroids required a longer duration of mechanical ventilation and longer ICU and hospital stay but had no statistically significant differences in ICU or hospital mortality (Supplementary Table S3).

Discussion

In our study, one third of subjects who remained mechanically ventilated following completion of a standard 10-day course of dexamethasone had evidence of rebound hypoxemia by day 7 following steroid cessation. These subjects had higher ICU mortality. Although there were no identifiable characteristics at time of steroid cessation to predict rebound hypoxemia, these subjects did have evidence of persistently elevated C-reactive protein by day 7.

In the overall cohort, steroid recommencement was common. Almost one third of subjects were commenced on an alternative corticosteroid, often within 2 days of completion of their standard 10-day dexamethasone course. These subjects had greater treatment intensity at time of steroid recommencement, had longer duration of mechanical ventilation and ICU stay, but no significant difference in mortality.

The prevalence of rebound hypoxemia in our cohort was much higher than reported in other COVID-19 studies (1 in 3 subjects compared with 1 in 10 in mixed severity cohorts^5,6). As we solely focused on ICU patients, this may suggest a higher likelihood of this phenomenon in sicker patients requiring invasive mechanical ventilation, compared with all patients receiving hospital-based care. Our finding is consistent with the non-COVID-19 ARDS literature, which describes a rebound inflammatory response with severe clinical relapse in one quarter to one third of subjects following abrupt corticosteroid cessation after 3 to 14 days in randomized controlled trials.¹² Furthermore, patients who abruptly cease and do not recommence corticosteroids may have worse outcomes with return to mechanical ventilation and higher mortality.^22,23 This specific complication of abrupt corticosteroid cessation appears to be consistent in both COVID-19 and non-COVID-19 patients with ARDS despite varying pathogenic mechanisms of lung injury.

Prior studies have suggested age, higher F_IO2 requirement, and higher inflammatory markers are independently associated with risk of rebound hypoxemia.⁶ Our study found subjects with rebound hypoxemia were older, by a median difference of 6 years, and had a C-reactive protein level more than double that of the stable cohort 1 week following steroid cessation, suggesting a persistent inflammatory state. Prolonged cytokine release syndrome and delayed viral clearance^13–15 have been implicated as causative factors for rebound hypoxemia seen in severe COVID-19 pneumonia. Our study provides further evidence to support the role of persistent inflammation with this phenomenon.

Organizing pneumonia has also been described as an important cause of rebound hypoxemia in multiple case reports.^8–10 However, although 1 in 5 subjects in our study had radiographic evidence of organizing pneumonia, we found no differences in rates of organizing pneumonia between subjects with stable or deteriorating P_aO2/F_IO2 or between subjects who ceased or recommenced steroids.

To date, randomized controlled trials have studied steroid durations of up to 14 days in COVID-19 ARDS.¹⁶ Although mortality benefit with corticosteroid use in critically ill patients with COVID-19 is well established, the ideal duration of therapy is still unclear. Current national and international guidelines suggest the use of intravenous or oral dexamethasone (or suitable alternative corticosteroid) for 7–10 days in this population.^17–19 However, our study has shown that a one-size-fits-all approach may be unwise. In particular, such an approach may be inadequate in a subset of mechanically ventilated patients with persistently high levels of inflammation. In some of these patients, clinicians decided to restart steroid therapy. Such practice has also been shown in another COVID-19 study,²⁰ although, in contrast, our subjects did not have higher mortality rates with greater steroid exposure. Use of antimicrobials in Remmington et al²⁰ is not reported but was high in our cohort. Recent evidence suggests there is a higher risk of superinfection with corticosteroid use in severe COVID-19 pneumonitis,²¹ and this should be considered as an important differential for rebound hypoxemia in this population. Adverse effects associated with prolonged corticosteroid use may also necessitate close observation of these patients.

Our study findings imply that, in critically ill subjects with COVID-19, rebound hypoxemia following steroid cessation was relatively common and associated with poorer outcomes. They also imply that clinicians should be aware of this important phenomenon when completing a standardized course of corticosteroid therapy for COVID-19 ARDS. Our findings further suggest that daily C-reactive protein measurement may assist clinicians in monitoring the trajectory of systemic inflammation following cessation of corticosteroids and that a persistently raised C-reactive protein may justify consideration of recommencing therapy. Further studies assessing the use of C-reactive protein and other inflammatory biomarkers to determine optimal duration of corticosteroid therapy are warranted. Importantly, recommencing steroids was not associated with higher mortality in our study.

Our observations suggest that clinical rather than radiological criteria appear to guide clinician decision-making to recommence steroid therapy, in particular, intensity of treatments such as neuromuscular blockade and prone positioning. The ideal duration of corticosteroids in COVID-19 ARDS, similarly to non-COVID-19 ARDS, remains unknown. The presence of persistent inflammation and ongoing need for ventilator and non-ventilator strategies to optimize oxygenation suggest a personalized approach is best utilized to determine corticosteroid duration rather than a standardized course. Current recommendations for non-COVID-19 ARDS support resumption of therapy for patients who rapidly deteriorate following abrupt cessation of corticosteroids,²³ and our study provides support for such practice in COVID-19 patients with ARDS also.

Our study adds to the limited data on the prevalence and factors associated with rebound hypoxemia following cessation of corticosteroids for critically ill patients with COVID-19 pneumonitis. We provided a detailed analysis of the oxygenation outcomes following completion of a standard 10-day course of dexamethasone and found that rebound hypoxemia in this population was associated with increased mortality. As international guidelines continue to only recommend 7- to 10-day courses of corticosteroids in this population, our findings have clinical relevance.

We acknowledge several limitations. This was a retrospective analysis with inherent limitations of selection bias and unmeasured confounders. We could not definitively exclude co-infection as a cause for rebound hypoxemia in this study. However, serial CT scans showed no differences in radiographic infective changes between groups. Furthermore, there were no differences in white cell count, antimicrobial administration, or positive blood cultures. A small proportion of subjects did not remain in the ICU beyond 48 hours following cessation of dexamethasone and oxygenation outcome data were unavailable for these subjects. However, given their stability and ICU discharge, it is unlikely they had significant rebound hypoxemia. Our cohort size was small, and outcomes of interest may be subject to type-2 error. Despite this, there remained clear signals from this study of associations between rebound hypoxemia, persistent inflammation, and increased mortality.

Conclusions

Rebound hypoxemia and steroid recommencement were relatively common occurrences in a cohort of mechanically ventilated COVID-19 subjects following completion of a standard 10-day course of dexamethasone. Such rebound hypoxemia was associated with higher C-reactive protein levels and higher mortality. Steroid recommencement was associated with longer duration of mechanical ventilation but no significant difference in mortality. These observations have clinical implications and can serve to guide decision-making around the duration of steroid therapy in patients with severe COVID-19 ARDS.

Author Disclosure Statement

There are no relationships to disclose for any of the authors.

Funding Information

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.