Early Treatment of Multisystem Inflammatory Syndrome in Children

Three children’s hospitals participated in the study: NewYork-Presbyterian Morgan Stanley Children’s Hospital, Cohen’s Children’s Hospital/Northwell Health, and NewYork-Presbyterian Komansky Children’s Hospital. Study researchers reviewed all MIS-C cases reported to local departments of health based on CDC criteria.¹⁶  Cases were included if they occurred in patients less than 21 years of age, met CDC criteria,¹⁶  had positive reverse transcription polymerase chain reaction or antibody testing for SARS-CoV-2, had no other plausible diagnoses, and presented between March 1, 2020 and April 1, 2021. Treatment in this study was based on local institutions’ protocols, standards, and clinician judgment. Cases were excluded if they received care in a nonparticipating hospital for more than 24 hours before transfer to any study site. Data were collected retrospectively from participating institutions’ electronic health records, deidentified, and stored in a secure database (REDCap, Vanderbilt University). Recorded data included patient demographics, illness features, clinical monitoring, cardiac and other radiographic imaging, laboratory values, treatments, and hospital course. Columbia University and local institutional review boards approved the study with a waiver of informed consent.

The primary outcome of the study was hospital LOS in subjects who received glucocorticoids within 48 hours of arrival to the treating hospital compared with receipt more than 48 hours after arrival. Arrival was defined as first presentation to the emergency department (for study sites) or first presentation to the study site itself (for transferred patients). Subjects who received IVIG but did not receive corticosteroids were included in the late treatment group. Subjects were excluded if they did not receive treatment with IVIG, because IVIG was considered standard of care within each hospital and no treatment likely reflected particularly mild presentation or atypical illness course. Secondary outcomes included ICU LOS, change in pediatric sequential organ failure assessment (pSOFA) score between initial presentation and 72 hours, and duration of vasoactive or inotropic infusion.¹⁷  The change in COVID-19 severity score between hospital arrival and 72 hours after arrival was used as an additional secondary outcome.¹⁸  This score was adapted from the inpatient portion of Kalil et al to incorporate vasoactive support and was scored as (1) not hospitalized, (2) hospitalized without organ support, (3) any noninvasive oxygen requirement, (4) vasoactive or inotropic medication, (5) invasive mechanical ventilation (IMV), (6) IMV + vasoactive or inotropic support, or (7) death (Supplemental Table 4). Subgroup analysis was completed for ICU patients and non-ICU patients, and for patients who received early high-dose glucocorticoid therapy, defined as greater than or equal to 10 mg/kg per dose methylprednisolone or equivalent.

Demographic, clinical, and laboratory characteristics of children who received early versus late or no glucocorticoids were compared. Continuous variables were reported as medians and interquartile ranges (IQR) and compared using Wilcoxon rank-sum tests. Categorical variables were reported as counts and percentages and compared using χ-squared tests or Fisher exact tests. Unadjusted comparisons based upon early versus late or no glucocorticoids were made for the primary outcome (hospital LOS) and all secondary outcomes listed above. The ICU LOS was similarly compared for subjects with ICU hospitalizations.

We then estimated the association between early glucocorticoids and hospital LOS using Poisson regression, modeling the number of hospital days as the outcome variable. The model was adjusted for variables with a P value < .1 in bivariate analysis, as well as those deemed important by the investigators (race and COVID-19 wave). Highly collinear or infrequent variables were not included. When composite variables incorporated other significant variables (such as pSOFA score incorporating the creatinine), only the composite variable was included. Missing data were multiply-imputed using predictive mean matching if at least 20% of data were present.¹⁹  Only laboratory measurements had more than 20% missing, and among these only 3 had more than 50%. Propensity scores were estimated for the multiply-imputed datasets using gradient-boosted machine regression,²⁰  and the data were inverse probability treatment weighted based upon these scores.²¹  Incidence rate ratios (IRR) and robust 95% confidence intervals (95% CIs) for hospital LOS were estimated for early glucocorticoids and all included covariates using these multiply-imputed, propensity-weighted datasets. Doubly-robust models were used because not all covariates were balanced after propensity weighting.²²  Imputation, propensity score weighting, and regression modeling are further detailed in the Supplemental Materials.

Secondary outcomes were modeled using linear, Poisson, and logistic regression for continuous, count, and binary outcomes, respectively. Models were adjusted for the same covariates as the primary analysis unless those covariates were composite with the secondary outcome. We performed several sensitivity analyses. We estimated the hospital LOS (1) only in subjects with ICU stays, (2) only in subjects who received any glucocorticoids, and (3) only considering treatment with glucocorticoids as high-dose glucocorticoids. We performed an additional sensitivity analysis investigating whether duration of fever preceding hospitalization might be associated with LOS and whether LOS might vary by hospital site. We used a general linear mixed model with a Poisson distribution, regressing hospital LOS on early glucocorticoids and all other previous covariates. For this model, we included duration of fever and hospital site ID as additional covariates. Hospital site ID was added as a random effect, and all other covariates were fixed effects. All secondary outcomes and sensitivity analyses were performed using the same imputations, propensity weights, and covariates as in the primary analysis, except in cases where a covariate was composite with the outcome.

Statistical significance was determined using an α of 0.05. All analyses were performed using R Version 4.1.3 (R Foundation for Statistical Computing).²³  Imputation was performed using the MICE package, and propensity score weighting with the MatchThem package.^24,25 

Subject enrollment is described in Fig 1. One hundred and fifty five patients were included in the initial cohort spanning April 17, 2020 to April 1, 2021. After exclusions, the final study cohort included 131 subjects, of whom 95 received early glucocorticoids. All subjects survived to discharge. Subject demographics, presenting symptoms, and clinical characteristics are described in Table 1 and Supplemental Table 5. Compared with subjects who received late or no glucocorticoids, subjects who received early glucocorticoids had lower mean arterial pressure (54 mmHg vs 62 mmHg, P = .003) within 24 hours of arrival, despite being similarly aged (9.0 vs 7.5 years, P = .084). Subjects who received early glucocorticoids were more likely to be admitted to the ICU (74% vs 31%, P < .001), require vasoactive support (49% vs 17%, P < .001) within the first 24 hours of arrival, and require respiratory support (P = .008) within the first 24 hours of arrival. Duration of fever preceding hospitalization was not associated with early glucocorticoids (5 vs 4 days, P = .3). Early glucocorticoid subjects were more likely to have abnormal laboratory findings in the first 24 hours from arrival, including higher C-reactive protein (CRP) (175 mg/L vs 86 mg/L, P < .001), higher N-terminal prohormone of brain natriuretic peptide (NT-proBNP) (4100 pg/mL vs 651 pg/mL, P = .008), and higher ferritin (516 ng/mL vs 208 ng/mL, P = .003). Early glucocorticoid subjects were also more likely to receive IVIG within 48 hours (89% vs 75%, P = .04).

Unadjusted outcomes after hospital arrival are reported in Table 2. When comparing initial values within 24 hours of arrival to subsequent values 72 hours after arrival, early glucocorticoid administration was associated with greater reduction in pSOFA score (−2 vs −1, P = .003) and more frequently had a reduction in COVID-19 severity score (31% vs 11%, P = .02) when compared with late or no glucocorticoid administration. Early glucocorticoid use was also associated with shorter ICU LOS (4 vs 9 days, P = .004), but not hospital LOS (5 vs 6 days, P = .5).

Results of the adjusted analysis are reported in Fig 2. Balance metrics after multiple imputation and propensity score weighting are presented in Supplemental Fig 3. Subjects who received early glucocorticoids had an adjusted IRR for hospital LOS of 0.75 (95% CI 0.58–0.96, P = .025) compared with subjects receiving late or no glucocorticoids, indicating association with a shorter hospital LOS. Administration of early IVIG was also associated with a shorter hospital LOS (IRR 0.56, 95% CI 0.34–0.93, P = .026). Initial pSOFA score was associated with a longer hospital LOS (IRR 1.07, 95% CI 1.01–1.13, P = .028). Asian American and Pacific Islander race, lower sodium levels, and elevated CRP were also associated with longer hospital LOS.

We modeled several secondary outcomes using the same strategy as in the main analysis. Early glucocorticoids were associated with improvement in the COVID-19 severity scale at 72 hours after arrival (odds ratio 8.85, P = .002), but early IVIG was not associated with an improvement in this score (P = .23) (Supplemental Table 6). Early glucocorticoids were not associated with change in pSOFA score from arrival to 72 hours after arrival, or with total time on vasoactive/inotropic medications (Supplemental Tables 7 and 8).

In a sensitivity analysis limited to patients who spent time in the ICU (n = 81, 62%), early glucocorticoids had a similar association with hospital LOS as in the main analysis (IRR 0.70, 95% CI 0.52–0.96, P = .027). Similar but nonsignificant associations with hospital LOS were observed for early IVIG and pSOFA score (Supplemental Fig 4). Early glucocorticoids demonstrated a similar association with ICU LOS, which did not reach statistical significance (IRR 0.76, 95% CI 0.55–1.05, P = .10) (Supplemental Fig 5). In a second sensitivity analysis limited to patients who ever received glucocorticoids (n = 111, 85%), early glucocorticoids were associated with an adjusted IRR for hospital LOS of 0.64 (95% CI 0.46–0.88, P = .006). Early IVIG demonstrated a similar but nonsignificant association with hospital LOS as in the main analysis (IRR 0.59, 95% CI 0.34–1.01, P = .053) (Supplemental Fig 6). Early high-dose glucocorticoids were not associated with hospital LOS (IRR 0.84, 95% CI 0.64–1.11, P = .22), although early IVIG remained associated with a shorter LOS in this model (IRR 0.53, 95% CI 0.33–0.86, P = .01) (Supplemental Fig 7).

Finally, we tested whether duration of fever preceding hospitalization and site ID changed the association of early glucocorticoids with hospital LOS. Both early glucocorticoids (IRR 0.8, 95% CI 0.7–0.92, P = .002) and early IVIG (IRR 0.63, 95% CI 0.51–0.77, P < .001) remained associated with shorter LOS. Longer fever duration was also associated with a shorter hospital LOS (IRR 0.9, 95% CI 0.86–0.94, P < .001). The pooled random intercept (representing the effect of each hospital site on the overall LOS) was 0.0006, indicating minimal between-site variability (Supplemental Fig 8).

This study reports an association between treatment with glucocorticoids within 48 hours of hospital arrival and faster hospital discharge in patients admitted with MIS-C, as evidenced by an associated 25% reduction in the expected hospital LOS after early compared with late or no glucocorticoids. In addition, we report an independent and stronger association between early treatment with IVIG and rate of hospital discharge, as evidenced by an associated 44% reduction in hospital LOS after early IVIG. These findings remained consistent after accounting for differences among patients related to demographics, seasonality, and severity of illness, and were reinforced by multiple sensitivity analyses. These findings support early administration of both glucocorticoids and IVIG upon diagnosis of MIS-C.

Few multi-institutional studies have explored associations between hospital LOS and timing of MIS-C therapy. Although the ideal treatment of MIS-C is unknown, many guidelines recommend a combination of IVIG and corticosteroids for treatment.^11,26  Previous observational studies showed an association with improved cardiac function, shorter ICU stays, and less treatment failure in patients who received IVIG and corticosteroids compared with IVIG alone.^12,13  The only randomized control trial found that the 2 therapies were equivalent with respect to hospital LOS, but did not study them in conjunction with each other.¹⁵  Our prior single-center study demonstrated an association between MIS-C protocol initiation and earlier treatment with IVIG and glucocorticoids, as well as shorter LOS, possibly pointing to the importance of early treatment in disease resolution.²⁷ 

Both glucocorticoids and IVIG have mechanistic explanations for halting the hyperinflammatory cascade in MIS-C, which may explain their association with shorter LOS in our study when given early. Glucocorticoids are potent immunomodulators, enacting nuclear transrepression of proinflammatory genes, resulting in inhibition of a diverse array of cytokines. Glucocorticoids have also been shown to have nongenomic effects on immune and endothelial cells, including membrane permeability, adenosine triphosphate production, and T-cell receptor signaling, which is hypothesized to result in a reduction in immune cell activity and inflammation.²⁸  Given the overwhelming hyperinflammatory state in MIS-C, it is possible that the multiple mechanisms of action of glucocorticoids are beneficial in controlling the patient’s inflammatory response. Although glucocorticoids are not universally effective in hyperinflammatory conditions, they are used and have shown benefit in other hyperinflammatory states similar to MIS-C, including severe COVID-19 in adults and high-risk KD patients.^29–31  This study provides further evidence that administering glucocorticoids earlier in the disease process of MIS-C may limit the inflammatory cascade that leads to tissue injury, organ dysfunction, and shock, and thereby lead to a faster recovery.

IVIG has multiple theorized mechanisms of action in MIS-C and KD, including neutrophil killing and subsequent interleukin-1-β inhibition, T-cell inhibition, and endothelial cell suppression.³²  Serum IgG in MIS-C patients has been shown to bind to cardiac microvascular endothelial cells.³³  It is possible that IVIG inhibits autoantibody production or directly neutralizes autoantibodies in MIS-C similar to other illnesses.³⁴  A recent observational study demonstrated equivalence in outcomes between corticosteroid monotherapy and combined treatment with corticosteroids and IVIG in MIS-C.³⁵  However, patients who received corticosteroid monotherapy in that study had higher rates of treatment failure related to cardiac dysfunction. One possible explanation for our report of a positive association of IVIG and outcomes is that our cohort had more patients with cardiac dysfunction and/or severe disease, and therefore experienced more benefit from IVIG, as all participating centers are tertiary or quaternary children’s hospitals capable of mechanical circulatory support.

High-dose glucocorticoid use has been studied without conclusive results in hyperinflammatory shock,¹²  but there was no association between early high-dose glucocorticoid treatment and improvement in therapeutic outcomes in our study. This may be because of the small number of patients who received high-dose treatment. Similarly, there was no association between early treatment with glucocorticoids or IVIG and ICU LOS or duration of vasoactive infusions, possibly because of the smaller size of the ICU subcohort. Early glucocorticoids were also not associated with improvement in the pSOFA score at 72 hours, which may be because of the study’s lack of mortality in this study. Serious adverse events were rare among patients receiving IVIG and/or glucocorticoids: 1 patient who was treated with IVIG and high-dose glucocorticoids developed gastrointestinal hemorrhage, whereas another with the same treatment developed steroid-induced psychosis. Both patients recovered from these events.

There are several limitations to this study. The study is retrospective and observational and therefore subject to confounding by indication. Patients who received earlier or higher-dose steroids were likely sicker (at least 1 hospital’s protocol dictated steroid dose based upon severity of illness). We attempted to mitigate this potential confound by using doubly-robust propensity score weighting in our analysis, with a machine-learning approach that is able to account for multiple levels of complex, nonlinear interactions among covariates. Balance was improved but not perfect after weighting. Although this study involves multiple children’s hospitals, all centers are in the New York City metropolitan area, which may introduce local selection bias of patient population or standards of practice. All sites were referral hospitals capable of advanced circulatory support, which may have selected for sicker patients. However, during the early months of the study period most small hospitals did not admit children because of pandemic limitations, which led to larger catchment areas for 2 of the study sites. Further, the understanding and approach to MIS-C changed during the study period, which extended from the beginning of the COVID-19 pandemic to the spring of 2021 and may have led to earlier patient presentation or initiation of therapy. Therefore, we included the temporal wave of COVID-19 as an adjustment term in our analysis. Finally, it is possible that patients infected with novel SARS-CoV-2 variants since the conclusion of the study period and presenting with MIS-C could display different phenotypic attributes and responses to therapy. Vaccines were not available for most subjects during the study period. It remains to be seen whether similar associations between early MIS-C treatment and hospital LOS will be observed in the postpandemic era now that most children have either been immunized against or recovered from SARS-CoV-2 infection.

Early treatment of MIS-C with glucocorticoids was associated with shorter hospital LOS after accounting for disease severity at presentation and other potential confounding factors. Early treatment with IVIG was also independently associated with shorter hospital LOS. These findings point to utility in administering both therapies soon after the diagnosis of MIS-C. Further study is needed to ascertain the optimal therapy for MIS-C.

We thank Dr Steven Kernie for his assistance in conceptualizing this study and mentorship in its implementation; and Dr Wei-Yann Tsai and Mr Jimmy Duong of the Columbia University Irving Institute for Clinical and Translational Research for their assistance in the study’s statistical design.

CDC

Center for Disease Control and Prevention

COVID-19

coronavirus disease 2019

CRP

C-reactive protein

IMV

invasive mechanical ventilation

IQR

interquartile ranges

IRR

incidence rate ratio

IVIG

intravenous immunoglobulin

KD

Kawasaki disease

LOS

length of stay

MIS-C

multisystem inflammatory syndrome in children

NT-proBNP

N-terminal prohormone of brain natriuretic peptide

pSOFA

pediatric sequential organ failure assessment

SARS-CoV-2

severe acute respiratory syndrome coronavirus 2