Immunomodulatory Therapy for MIS-C

We conducted a systematic review and meta-analysis of individual participant data. The protocol of this study was registered on PROSPERO (CRD42021292162) before collecting and analyzing data, and the reporting followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (Appendix 1).¹⁸  The systematic review aimed at identifying potential other studies fulfilling our eligibility criteria that were not included in the International MIS-C Treatment Collaborative.

On February 2, 2022, we systematically searched Medline, Embase, Cochrane Central Register of Controlled Trials (CENTRAL) and Web of Science for articles published from January 1, 2020 to February 1, 2022, with no language restriction. An update of this literature search was performed on January 31, 2023. We developed a search algorithm based on a combination of terms related to “MIS-C,” “treatments” (intravenous immunoglobulins and/or glucocorticoids), and “child.” The full search algorithm is presented in Supplemental Table 3. We completed this search by screening the reference lists of all included articles.

Studies were eligible if: (1) they were designed as randomized or nonrandomized quasi-experimental or observational comparative therapeutic studies, (2) the objective was to formally compare outcomes of children treated with IVIG plus glucocorticoids, glucocorticoids alone, or IVIG alone as initial therapy, and (3) the patient population fulfilled World Health Organization (WHO),¹⁹  Centers for Disease Control and Prevention (CDC)²⁰  or Royal College of Pediatrics and Child Health criteria for MIS-C²¹  (details in Appendix 2).

Exclusion criteria included: (1) noncomparative studies (e.g., cases series, single-arm cohorts, or 2-arm studies with less than 10 patients per arm), (2) nonoriginal studies (e.g., review, systematic review), (3) studies with duplicate patients, (4) studies that did not compare 1 of the 3 initial therapies (IVIG plus glucocorticoids, glucocorticoids alone, or IVIG alone, and (5) in vitro studies.

Two independent reviewers (N.O. and E.V.) screened the titles and abstracts of all publications identified by the search strategy and examined the full text of any potentially eligible article. Disagreements were resolved by discussion with a third author (F.A.) to reach consensus.

Anonymized individual patient data from the 3 consortia were collected in a secure database. Collected data included baseline characteristics needed to develop the propensity score and outcomes.

The risk of bias and quality of evidence assessment followed the previously published living WHO guidance for clinical management of MIS-C, which provided an independent assessment of the studies included in this meta-analysis.¹⁵  The risk of bias assessment followed the Cochrane risk-of-bias tool for nonrandomized studies of interventions (ROBINS-I),²²  which contains 7 items: confounding, selection of participants into the study, classification of the intervention, deviations from intended interventions, missing data, measurements of outcomes, and selections of the reported result. The certainty of evidence was assessed using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) tool recommended for meta-analysis.^23,24  The certainty of evidence was rated for each outcome as “high,” “medium,” “low,” or “very low.”

The primary outcome was cardiovascular dysfunction on or after day 2 of initial therapy, defined as either a left ventricular ejection fraction (LVEF) < 55% or the use of vasoactive or inotropic amine. As previously published, this composite cardiovascular outcome was chosen because low LVEF does not always result in the use of vasopressors, and distributive shock requiring hemodynamic support does not consistently accompany low LVEF in MIS-C.¹³ 

The day the first immunomodulatory treatment was administered was considered day 0. Treatment combination (IVIG plus glucocorticoids) was considered initial therapy when the beginning of administration of the 2 therapies occurred within 1 calendar day. Based on this, 3 treatment groups were analyzed for the primary outcome: IVIG alone, IVIG plus glucocorticoids, and glucocorticoids alone.^12–14 

• Hemodynamic support on or after day 2 of initial therapy, defined as the use of vasoactive or inotropic amine;

• LVEF< 55% on or after day 2 of initial therapy, through discharge;

• Ventilatory support (invasive or noninvasive) on or after day 2 of initial therapy, through discharge;

• Fever on or after day 2 of initial therapy, through discharge; and

• Second-line therapy, also known as secondary treatment, including glucocorticoids, second doses of IVIG, or biologic medications prescribed for MIS-C at least 24 hours after the initial therapy.

• Patients fulfilling WHO versus CDC criteria for MIS-C;

• Patients with or without initial cardiovascular dysfunction (defined as either initial hemodynamic support or initial LVEF< 55%);

• Patients with or without initial hemodynamic support (defined as the use of vasoactive or inotropic amine);

• Patients with or without initial LVEF < 55%;

• Children < or ≥6 years of age;

• Patients with or without criteria for complete KD;

• Patients with or without lower respiratory tract symptoms.

The detailed definition of each subgroup is provided in Appendix 3. Because of sample size limitation, these subgroup analyses were only conducted for the IVIG plus glucocorticoids versus IVIG alone comparison. The age dichotomization was based on a previously published study.¹⁴ 

The primary analysis used propensity score matching.^12–14  The propensity score was calculated with a multivariable logistic regression model, allowing to estimate for each patient the probability of receiving 1 of the initial therapies according to baseline characteristics. The baseline characteristics used to build the propensity score were: continent, age, sex, comorbidities (defined by any chronic condition), obesity, gastrointestinal symptoms, lower respiratory tract symptoms, neurologic symptoms, criteria for complete KD,²⁵  intensity of inflammatory response (C-reactive protein level > or ≤150 mg/L), initial PICU care, and initial hemodynamic support. All baseline characteristics were considered at admission, ie, before or on the day of initial therapy. The detailed definition of each variable is provided in Appendix 3.

Patients from each treatment group were matched by their propensity score using nearest-neighbor matching without replacement, with a minimum caliper of 0.2.²⁶  The ratio was 1 patient from IVIG plus glucocorticoids group matched with 1 patient receiving IVIG alone. Given the lower number of cases in the glucocorticoids alone group, the ratio was 1 patient from glucocorticoids alone group matched with 2 patients receiving IVIG alone and with 2 patients receiving IVIG plus glucocorticoids. The balance between the treatment groups for each covariate was assessed with a standardized difference less than 0.1 considered acceptable.¹⁰  Individual participant data meta-analyses were conducted as a 1-stage approach of binary outcomes via conditional logistic regression analysis, using the matched cohorts to test the association between each treatment groups and outcomes.¹⁶  Findings were expressed as odds ratios (ORs), and 95% confidence intervals (CIs). The analysis also adjusted for continent and for heterogeneity between studies by using random-effects modeling.²⁷  As a measure of between study heterogeneity, the between study variance was provided for each outcome.^28,29 

Based on previous publications,^12–14  a low proportion of missing data were expected. Thus, a complete-case analysis was conducted.

• First, the data were analyzed using inverse probability of treatment weighting (IPTW), which is an alternative approach to propensity score matching to account for indication bias in nonrandomized design.^10,11  Unlike propensity score matching, this strategy has the advantage of including all patients in the analysis¹¹ ;

• Second, a propensity score-matched analysis was performed with fixed effect for continent and study;

• Third, propensity score-matched analysis was performed with a minimum caliper of 0.1, to explore potential remaining unbalance between groups;

• Fourth, a propensity score-matched analysis was conducted with within-study matching to account for potential heterogeneity between studies. Indeed, a recent study compared within and across-study matching for individual participant data meta-analysis and found that across matching reduced the risk of bias if the prevalence of treatment was similar across studies.²⁸  Another similar study suggested that across-study matching may improve covariate balance.³⁰  In our meta-analysis, the prevalence of IVIG plus glucocorticoids was similar across studies, ranging from 52% to 69% of cases. This led us to choose matching across studies for the main analysis. However, as within study matching may offer other advantages,²⁸  and to explore the influence of this approach on the results, a sensitivity analysis using within study matching was also conducted;

• Fifth, a propensity score–matched analysis was conducted with double adjustment on likely confounding variables: initial hemodynamic support and initial LVEF < 55%.¹²  This strategy has been proposed to remove residual confounding after propensity score matching for the main potential confounders³¹ ;

• Sixth, an IPTW analysis was conducted with double adjustment on initial hemodynamic support and initial LVEF < 55%³¹ ;

• Seventh, a propensity score–matched analysis was conducted with initial left ventricular dysfunction considered on day 0 or day 1 included in the propensity score, as some patients only had an initial echocardiography on day 1;

• Eighth, a propensity score-matched analysis was conducted with fever duration before first-line therapy as an additional baseline variable to account for potential differences between groups in the delay between disease onset and initial treatments¹² ;

• Ninth, a logistic multivariable regression analysis adjusted on the variables included in the propensity score was conducted;

• Tenth, for comparisons of glucocorticoids alone versus IVIG alone and glucocorticoids alone versus IVIG plus glucocorticoids, a propensity score-matched analysis using a 1:1 matching was also conducted.

All statistical tests were 2 sided, with P < .05 considered statistically significant. Statistical analyses were conducted using using R v4.1.1 (http://www.R-project.org/).

Among the 2635 studies retrieved from Medline, Embase, CENTRAL and Web of Science, 918 duplicates were removed, and 1717 studies were screened by title and abstract. Among them, 76 were potentially eligible. After full text examination, 3 studies were finally included (see detailed PRISMA diagram Fig 1A). The update of the literature search on January 31, 2023 retrieved 628 additional studies, among which a single center retrospective cohort met our inclusion criteria,³²  but could not be included in our individual participant data meta-analysis (individual data not available).

These 3 studies were nonrandomized cohorts. Thus, despite the use of propensity score analysis in all studies, the risk of bias for confounding using the ROBINS-I tool was classified as serious for all included studies. As a result, the level of certainty was very low for all GRADE outcomes (Appendix 4).

Baseline characteristics of the 958 included children are detailed in Table 1. The median age was 8.1 years, interquartile range (4.2–11.7), sex ratio was 0.7 (female to male), without major differences across studies. Incidence of initial PICU care was 34% (161 of 468) in the BATS study, 45% (158 of 349) in the Overcoming COVID-19 study, and 68% (72 of 106) in Pandor study. Rate of initial LVEF < 55% followed the same trend (23% (61 of 267) in BATS study, 30% (89 of 295) in Overcoming COVID-19 study, and 47% (50 of 106) in Pandor study (Table 1).

The rate of cardiovascular dysfunction on or after day 2 was quite high in children who already had initial cardiovascular dysfunction compared with children without initial cardiovascular dysfunction (46%, 149 of 327 versus 9%, 34 of 378, respectively). Similar observations were seen for hemodynamic support on or after day 2, LVEF < 55% on or after day 2, and ventilatory support on or after day 2 (Supplemental Table 4).

Among the 958 included children with MIS-C, 482 were in the IVIG alone group, 387 in the IVIG plus glucocorticoids group, and 89 in the glucocorticoids alone group (Fig 1B, Supplemental Table 5). No included children had received COVID-19 vaccines because of the timing of the study enrollment period. No child received remdesevir or other antiviral therapy.

For the main analysis comparing IVIG plus glucocorticoids versus IVIG alone, 311 children were matched in each group. After matching, the balance between the treatment groups for baseline characteristics was satisfactory (Supplemental Fig 3, A and B). The primary outcome (cardiovascular dysfunction on or after day 2) occurred in 58 of 311 (19%) matched children from IVIG plus glucocorticoids group versus 85 of 311 (27%) in the matched IVIG alone group. Patients treated with IVIG plus glucocorticoids had significantly less cardiovascular dysfunction on or after day 2 as compared with those treated IVIG alone as initial therapy (OR = 0.62 [0.42–0.91], P = .014).

All secondary outcomes favored the IVIG plus glucocorticoids group, except for ventilatory support on or after day 2 (Fig 2A, Supplemental Table 6A). Similarly, all sensitivity analyses showed improved outcomes for children treated with IVIG plus glucocorticoids compared with IVIG alone, including IPTW and within-study propensity score-matched analysis (Supplemental Table 7A). Subgroup analyses also favored IVIG plus glucocorticoids group, except for those children with initial lower respiratory tract symptoms (OR = 0.77 [0.46–1.30], P = .33, Table 2) or full criteria for KD (OR = 0.59 [0.19–1.83], P = .36).

For the main analysis, 75 children from the glucocorticoids alone group, who all came from the BATS study, were matched to 150 children from IVIG alone group. The balance between the matched treatment groups for baseline characteristics was satisfactory (Supplemental Fig 3, C and D). The primary outcome of cardiovascular dysfunction on or after day 2 occurred in 10 of 75 (13%) matched children from the glucocorticoids alone group and in 33 of 150 (22%) in the IVIG alone group, with no significant difference between the groups (OR 0.57 [0.31–1.05], P = .07, reference: IVIG alone group).

All secondary outcomes analyses showed no significant association between treatment with glucocorticoids alone versus IVIG alone and outcome, except for LVEF < 55% on or after day 2 (OR 0.13 [0.03–0.59], P = .008, Fig 2B, Supplemental Table 6B). Sensitivity analyses provided similar nonsignificant results, with the exception of a propensity score-matched analysis that included fever duration before first-line therapy which favored glucocorticoids alone (OR 0.53 [0.28–0.99], P = .046, reference: IVIG alone group, Supplemental Table 7B).

For the main analysis, 54 children from the glucocorticoids alone group were matched to 108 children from IVIG plus glucocorticoids group. The balance between the matched treatment groups for baseline characteristics was satisfactory (Supplemental Fig 3, E and F). The primary outcome of cardiovascular dysfunction on or after day 2 occurred in 8 of 54 (15%) children in the glucocorticoids alone group versus 23 of 108 (21%) in the IVIG plus glucocorticoids group, with no statistically significant difference between groups (OR 0.67 [0.24–1.86], P = .43, reference: IVIG plus glucocorticoids group).

In the secondary outcome analyses, the glucocorticoids alone group was associated with a higher rate of second line therapy (OR 6.95 [3.73–12.95], P < .0001) and persistent fever on or after day 2 (OR 2.16 [1.18–3.93], P = .012, Fig 2C, Supplemental Table 6C). All sensitivity analyses also found no significant association between treatment and the primary outcome (Supplemental Table 7C).

In a meta-analysis of patient-level data from nearly 1000 participants enrolled in 3 observational studies across 5 continents, we found that initial treatment of MIS-C with IVIG plus glucocorticoids was associated with improved cardiovascular dysfunction when compared with IVIG alone. These findings were consistent across sensitivity analyses, secondary outcomes, and subgroups analyses. By contrast, initial treatment with glucocorticoids alone was not associated with improved cardiovascular dysfunction, compared with IVIG alone or glucocorticoids plus IVIG.

The findings regarding glucocorticoids plus IVIG versus IVIG alone diverged in the 3 original studies, as the BATS study found no significant difference in outcomes,¹⁴  whereas Pandor and Overcoming COVID-19 studies showed IVIG plus glucocorticoids to be associated with superior outcomes.^12,13  Because these 3 studies were similar in their statistical approach to reduce confounding by indication,^12–14  the statistical approach used seems unlikely to explain divergent findings. However, the initial severity of the illness varied substantially across studies.^33,34  The Pandor and Overcoming COVID-19 studies enrolled more children with initial cardiovascular dysfunction requiring PICU admission,^33,34  and these patients were as at higher risk of persistent organ dysfunction on or after day 2 than those without in our analysis. Based on this, 2 hypotheses could be offered: (1) the benefit of combination therapy could be restricted to the most severely ill children, and (2) the benefit of combination therapy could be present both in severe and mild disease, but inclusion of milder MIS-C may reduce the power to detect differences as these outcomes are less frequent. Our subgroup analyses found that the benefit of the combination therapy compared with IVIG alone was substantial both in children with or without initial cardiovascular dysfunction, suggesting that the benefit of combination therapy may not vary according to initial severity. However, a recent study addressing this question suggested that the most important differences between BATS and Overcoming COVID-19 studies was initial cardiac involvement, PICU care, and vasopressors use.³⁵  Taken together, these findings suggest that inclusion of children with milder disease in the BATS study may have reduced the power to detect significant differences in populations where outcomes are rare and may have driven the divergent findings of these 3 original studies.³⁵  It may also indicate that for some children with mild initial illness, a monotherapy with glucocorticoids may be appropriate.

A third source of heterogeneity may be the different primary outcomes selected. Indeed, the Pandor study used persistent fever as the primary outcome, with organ dysfunction being a secondary outcome.¹²  Overcoming COVID-19 studied a primary outcome of cardiovascular dysfunction, as defined in this meta-analysis,¹³  whereas the BATS study used a composite outcome (hemodynamic and/or ventilatory support, or death).¹⁴  Of note, in this meta-analysis, ventilatory support was the only secondary outcome without significant difference between IVIG plus glucocorticoids and IVIG alone groups. This should be put in perspective with accumulated evidence suggesting that severe pulmonary involvement may not be a key feature of MIS-C and is not part of the WHO or the new CDC definitions.^19,36  Cardiovascular dysfunction may represent a more specific outcome and may be preferred in future therapeutic studies.¹³  A recently published single center study included 68 children treated with glucocorticoid monotherapy and compared them to children treated with IVIG plus glucocorticoids.³²  The primary outcome in that study was failure of initial therapy as evidenced by fever, worsening or lack of improvement of laboratory, cardiac, or clinical factors. They found no significant differences across treatment groups, but comparison with our findings is limited as this study did not compare IVIG alone versus IVIG plus glucocorticoids, and the composite outcome differed from ours.³²  This underlines the need for harmonizing primary outcomes across therapeutic studies to allow comparisons and meta-analyses.

The main strengths of our meta-analysis include the large number of patients and the individual participant data analysis. This offered several advantages.²⁹  First, it allowed applying the propensity score method. Indeed, as all included studies were not randomized trials, limiting the risk of bias was critical. The baseline variables used to calculate the propensity score varied across the original studies and outcomes analyzed also differed. Thus, using individual participant data were the only way to calculate a homogeneous propensity score for all included children, allowing adequate balance between groups for baseline covariates and to harmonize outcomes.²⁹  Second, it allowed conducting sensitivity analyses not present in original studies, reinforcing the robustness of our findings.²⁹  It also allowed conducting subgroup analyses, which were not possible in original studies because of sample size limitations.

Several limitations were present. First, because of the nonrandomization, the certainty was very low for all GRADE outcomes. Despite the propensity score approach achieving adequate balance for baseline covariates, we can’t exclude remaining confounding by indication because of unmeasured covariates. Second, as MIS-C diagnosis relies on clinical criteria, misdiagnosis can’t be excluded. Additionally, echocardiographic findings as a clinical criterion were not independently reviewed in all cohorts once sent to the coordinating centers. Subgroup analyses by the degree of initial cardiac dysfunction may require future studies. Third, as most children were enrolled in the respective studies in 2020, MIS-C triggered by Δ and ο variants was not included. Fourth, potentially evolving treatment norms may have influenced outcomes. However, as most MIS-C cases included in this meta-analysis occurred early in the pandemic, this hypothesis seems unlikely. Fifth, the number of children treated with glucocorticoids alone as initial therapy was low, which may have reduced the power to detect differences for this group and may have increased the risk of type 2 error. As the primary outcome analysis for glucocorticoids alone versus IVIG alone trended toward significance, differences may emerge with increasing the number of participants in future analyses. Sixth, as the pathophysiology of MIS-C remains unclear, the biology underlying the benefit of combination therapy versus IVIG alone requires further investigation. Seventh, the systematic review considered articles published up to February 1, 2022. An update of the literature on January 31, 2023 identified the above-mentioned single center retrospective cohort meeting our inclusion criteria that was not included in our individual participant data meta-analysis.³²  Eighth, different healthcare resources or other treatments depending on WHO region may have influenced outcomes. However, children that received cytokine blockers as initial therapy were not included. No included patient received antiviral therapy. No included children received a COVID-19 vaccine.

While awaiting results from randomized trials, this meta-analysis offers further support for initial treatment with IVIG plus glucocorticoids as compared with IVIG alone for children who meet criteria for MIS-C, especially for severely ill patients. As the primary outcome analysis found no significant difference between treatment with glucocorticoids alone versus IVIG alone or versus IVIG plus glucocorticoids, monotherapy with glucocorticoids may be considered, especially in settings of limited IVIG availability. However, caution should be taken when using this initial treatment, as exploratory secondary outcome analysis suggested better outcomes for combination therapy compared with glucocorticoids alone.

In a meta-analysis of nonrandomized trials of MIS-C treatments, IVIG plus glucocorticoids was associated with improved cardiovascular dysfunction compared with IVIG alone. In contrast, glucocorticoids alone were not significantly associated with improved cardiovascular dysfunction compared with immunoglobulins alone or compared with immunoglobulins plus glucocorticoids. Secondary outcome analyses, that should be considered exploratory, suggested better outcomes associated with glucocorticoids alone compared with IVIG alone and better outcomes associated with glucocorticoids plus IVIG compared with glucocorticoids alone.

We thank all investigators that contributed to 1 of the 3 consortia. We are grateful to Santé Publique France, Société Française de Pédiatrie, Groupe de Pediatrie Generale, Groupe de Pathologie Infectieuse Pédiatrique, Groupe Francophone de Réanimation et d’Urgences Pédiatriques, Société Française de Cardiologie, Filiale de Cardiologie Pédiatrique et Congénitale, Société Francophone Dédiée à L’étude des Maladies Inflammatoires Pédiatriques, and Filière de Santé des Maladies Auto-immunes et Auto-inflammatoires Rares for their participation in the French Covid-19 Pediatric Inflammation Consortium study. We thank Dr Corinne Levy, Pr Robert Cohen, Stéphane Béchet, Isabelle Ramay, BSc, Claire Prieur, BSc, Marine Borg, Aurore Prieur, BSc, Laura Meyet, LLM, Jéremy Levy, BSc, Stéphane Bechet, MSc, and Sofia Abbou, LLM, from ACTIV (Association Clinique et Thérapeutique Infantile du Val-de-Marne), Créteil, France; Cecile Hoffart, MSc, and Maxime Brussieux, BSc, from Clinical Research Center, Centre Hospitalier Intercommunal de Créteil, Créteil, France; Daniel Levy-Bruhl, MD, Mireille Allemand, Scarlett Georges, BSc, Valerie Olie, PhD, Nolween Regnault, PhD, and Jerome Naud, PharmD, from Santé Publique France, Agence Nationale de Santé Publique, Saint-Maurice; Murielle Herasse, PhD, from Filière de Santé Des Maladies Auto-immunes et Auto-inflammatoires Rares (FAI2R), Lyon, France.

CD

cardiovascular dysfunction

CI

confidence intervals

GRADE

Grading of Recommendations Assessment, Development, and Evaluation

IPTW

inverse probability of treatment weighting

IVIG

intravenous immunoglobulin

LVEF

left ventricular ejection fraction

MIS-C

multisystem inflammatory syndrome in children

OR

odds ratio