This two-center retrospective cohort study was conducted at Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, and the First Affiliated Hospital of Zhengzhou University. The study complied with the Declaration of Helsinki and was approved by the institutional review boards of both participating centers (Tongji Hospital approval No. TJ-IRB202604153; First Affiliated Hospital of Zhengzhou University approval No. 2025-KY-1274-001). Written informed consent was obtained from all patients or their legal representatives. The study was registered at ClinicalTrials.gov (NCT03268642).

Consecutive patients admitted with suspected FM between January 2020 and January 2025 were screened. FM was adjudicated according to the Chinese expert consensus and the American Heart Association scientific statement. Required features were an abrupt presentation with severe heart failure, cardiogenic shock, hypotension requiring vasoactive or inotropic support, or life-threatening arrhythmia; biochemical evidence of myocardial injury; and exclusion of alternative causes of acute myocardial injury or circulatory collapse. Acute coronary syndrome, sepsis-related myocardial injury, hypovolemic shock, and pre-existing structural myocardial disease were excluded using coronary angiography or coronary computed tomography angiography when feasible, serial electrocardiography and biomarker trajectories, echocardiography, cardiac magnetic resonance imaging when clinically possible, and multidisciplinary adjudication.

Of 404 patients screened, those aged < 18 years, those with an admission left-ventricular ejection fraction (LVEF) > 50%, and those with a confirmed alternative diagnosis were excluded. The final cohort comprised 316 adults, categorized according to IVIG exposure during hospitalization as IVIG-treated (n = 184) or non-IVIG-treated (n = 132) [Figure 1].

All patients received comprehensive care according to hemodynamic status, organ perfusion, arrhythmia burden, and treating-team judgment, including vasoactive or inotropic therapy, mechanical ventilation, renal replacement therapy, and temporary MCS when clinically indicated. Intra-aortic balloon pump and peripheral VA-ECMO were used for refractory cardiogenic shock or persistent hypoperfusion despite pharmacological support.

IVIG was prescribed at the treating team's discretion, without a study-mandated dose or schedule. Both centers used the same 5% human immunoglobulin preparation (Human Immunoglobulin [pH4] for Intravenous Injection; Hualan Bio, Xinxiang, China). Actual daily doses ranged from 5 to 20 g and treatment duration from 1 to 5 days. Cumulative exposure was categorized as < 10 g (n = 65), 10 to < 20 g (n = 83), or ≥ 20 g (n = 36), representing low, intermediate, and higher observed exposure, respectively. IVIG was not administered according to a fixed weight-based protocol; weight-adjusted exposure was calculated retrospectively. Treatment timing was defined as the interval from hospital admission to the first IVIG dose, with initiation within 4 days classified as early and initiation after 4 days as delayed; 3- and 5-day thresholds were examined in sensitivity analyses.

Methylprednisolone was administered to all 184 IVIG-treated patients and 60 of 132 non-IVIG-treated patients. Intravenous methylprednisolone was given at 200-500 mg/day for 3-5 days and subsequently tapered according to clinical response, using the same overall treatment framework at both centers. Neuraminidase inhibitors and empirical antibacterial therapy were administered when clinically indicated and were not determined by IVIG exposure. Detailed information regarding IVIG formulation, infusion procedures, weight-adjusted exposure, safety monitoring, corticosteroid tapering, and antiviral and antibacterial regimens is provided in Supplementary Table 1.

Demographic characteristics, presenting symptoms, admission vital signs, laboratory measurements, imaging findings, organ-support therapies, and in-hospital outcomes were abstracted from electronic medical records using a standardized case report form. Baseline variables included age, sex, blood pressure, cardiac troponin I (cTnI), N-terminal pro-B-type natriuretic peptide (NT-proBNP), white blood cell count, high-sensitivity C-reactive protein (hsCRP), alanine aminotransferase, aspartate aminotransferase, creatinine, circulating cytokines, and admission LVEF. Center, admission year, symptom-to-admission interval, interhospital transfer, residence, insurance/payment category, distance to the tertiary center, admission lactate, Sequential Organ Failure Assessment score, vasoactive-inotropic score within 24 h, and MCS initiated within 24 h were collected to characterize healthcare access, treatment selection, and early disease severity. Serial echocardiographic and laboratory data were also abstracted for descriptive analyses of admission, nadir, and discharge LVEF and in-hospital cardiac and inflammatory biomarker profiles.

The primary outcome was all-cause in-hospital mortality. Secondary outcomes were acute hepatic or renal dysfunction and successful ECMO weaning with survival to discharge. Acute hepatic or renal dysfunction was a composite of acute liver injury (ALI) and/or Kidney Disease: Improving Global Outcomes (KDIGO)-defined acute kidney injury (AKI). ALI was defined as persistent elevation of alanine aminotransferase or aspartate aminotransferase to > 3 times the upper limit of normal. All patients requiring continuous renal replacement therapy (CRRT) were classified as having AKI. For cumulative-incidence analyses of ALI/AKI, the first occurrence of ALI/AKI was the event of interest; in-hospital death before ALI/AKI and discharge alive without prior ALI/AKI were treated as competing events. Successful ECMO weaning with survival to discharge was defined as ECMO decannulation followed by survival to hospital discharge. For ECMO time-to-event analyses, time zero was the date of ECMO cannulation, and the event time was the date of ECMO decannulation in patients who subsequently survived to discharge. Death before qualifying for successful weaning was treated as a competing event. A patient who was decannulated but subsequently died before discharge would not be classified as having achieved the endpoint and would instead be coded as a competing death at the date of death. In the present cohort, no patient underwent heart transplantation before ECMO weaning, was transferred or discharged while receiving ECMO, or died after ECMO decannulation but before hospital discharge.

Histopathological analyses were restricted to clinically indicated endomyocardial biopsy specimens obtained during hospitalization; tissue acquisition was not mandated by the study protocol. Specimens were eligible when sufficient myocardium was available for hematoxylin-and-eosin staining and immunohistochemical assessment and when tissue processing, staining, and image quality permitted reliable quantification. Specimens with insufficient myocardial tissue, substantial processing artifact, or inadequate staining were excluded.

Eleven patients met these criteria, including 6 IVIG-treated and 5 non-IVIG-treated patients. Formalin-fixed, paraffin-embedded sections (4 μm) were stained with hematoxylin and eosin. Immunohistochemistry was performed using antibodies against myeloperoxidase (MPO; Abcam, ab109116, clone EPR4792, 1:250), CD68 (Abcam, ab125157, 1:100), CD3 (Abcam, ab17143, clone F7.2.38, 1:20), and CD20 (Abcam, ab78237, clone EP459Y, 1:50) to identify neutrophils, macrophages, T cells, and B cells, respectively. Five non-overlapping myocardium-containing high-power fields per marker were evaluated for each specimen, excluding endocardium, thrombus, and areas with major processing artifact. Positive cells were quantified as cells/mm² using QuPath version 0.4.4. Two pathologists, blinded to treatment allocation, independently reviewed the specimens and quantitative results.

Continuous variables are presented as median (Q1, Q3) and were compared using the Mann-Whitney U test. Categorical variables are presented as n (%) and were compared using the chi-square test or Fisher's exact test, as appropriate. Variable-level missingness is reported in Supplementary Table 2; no listed variable had > 5% missing observations. Missing covariate values were handled using multiple imputation by chained equations with 20 imputed datasets, and estimates were pooled using Rubin's rules.

For all-cause in-hospital mortality, Kaplan-Meier curves were compared using log-rank tests, and multivariable Cox proportional-hazards models were used for adjusted analyses. Time zero was hospital admission, and patients discharged alive were censored on their discharge dates. Numbers at risk were generated from the same survival objects used to estimate the Kaplan-Meier curves. IVIG initiation was additionally modeled as a time-varying exposure to address immortal-time bias. Proportional-hazards assumptions were assessed using Schoenfeld residuals.

For ALI/AKI, cumulative incidence functions were estimated using the Aalen-Johansen method and compared using Gray's test, with in-hospital death before ALI/AKI and discharge alive without prior ALI/AKI treated as competing events. Adjusted associations were estimated using cause-specific Cox proportional-hazards models. Because methylprednisolone use differed substantially between treatment groups, a prespecified analysis was restricted to methylprednisolone-treated patients. Propensity-score overlap weighting was additionally performed in this cohort, with covariate balance assessed using absolute standardized mean differences; values < 0.10 were considered acceptable.

For successful ECMO weaning with survival to discharge, cumulative incidence functions were estimated using the Aalen-Johansen method and compared using Gray's test, with death before qualifying for successful weaning treated as a competing event. Time zero was ECMO cannulation. The Fine-Gray subdistribution hazards model was used for the primary adjusted analysis, with cause-specific Cox, accelerated failure-time, and penalized logistic models used as sensitivity analyses. Exact model-specific covariates, censoring rules, competing-event definitions, and analytic specifications are summarized in Supplementary Table 3.

Dose- and timing-related analyses were exploratory. A day-4 landmark analysis evaluated subsequent mortality according to cumulative IVIG exposure, and alternative early-treatment thresholds of 3, 4, and 5 days were examined. Clinical subgroup analyses were performed according to white blood cell count, alanine aminotransferase, aspartate aminotransferase, and creatinine. All tests were two-sided, with P < 0.05 considered statistically significant. Analyses were performed using R version 4.2.2.

Among 404 patients screened for suspected FM, 316 met the eligibility criteria; 184 received IVIG, and 132 did not [Figure 1]. The cohort included 149 patients from Tongji Hospital and 167 from the First Affiliated Hospital of Zhengzhou University. The median age was 38.00 years (Q1, Q3: 25.00, 53.00), and 157 patients (49.68%) were male [Table 1].

Most admission vital signs, cardiac and inflammatory biomarkers, organ-function indices, cytokines, and admission LVEF did not differ significantly between groups. Nevertheless, treatment allocation was clearly nonrandom: IVIG-treated patients were admitted in more recent years, had a shorter symptom-to-admission interval, more often had urban employee or commercial insurance, and had higher admission lactate, 24-h vasoactive-inotropic scores, and more frequent MCS use within 24 h [Table 1]. Coronary angiography or coronary computed tomography angiography was performed in 287 patients, and cardiac magnetic resonance imaging was performed in 122, of whom 103 had Lake Louise-compatible findings. Missingness was < 5% for each listed variable [Supplementary Table 2].

In-hospital management differed substantially between groups [Table 2]. Methylprednisolone was administered to all 184 IVIG-treated patients and to 60 of 132 non-IVIG-treated patients. MCS was used in 145 IVIG-treated and 35 non-IVIG-treated patients, and intra-aortic balloon pump, ECMO, temporary pacing, CRRT, and vasoactive agents were generally used more frequently in the IVIG group.

In-hospital death occurred in 17 IVIG-treated patients (9.24%) and 51 non-IVIG-treated patients (38.64%). The ALI/AKI composite occurred in 53 IVIG-treated patients (28.80%) and 72 non-IVIG-treated patients (54.55%) [Table 2].

The Kaplan-Meier curves separated early and remained divergent through discharge [Figure 2A]. In the primary multivariable Cox model, IVIG use was associated with a lower hazard of in-hospital death (hazard ratio [HR], 0.23; 95% confidence interval [95%CI]: 0.14-0.40; P < 0.001). Modeling IVIG initiation as a time-varying exposure yielded a similar association (HR, 0.15; 95%CI: 0.08-0.29; P < 0.001) [Table 3].

The methylprednisolone-treated cohort comprised 184 IVIG-treated and 60 non-IVIG-treated patients, with 17 and 18 deaths, respectively. Survival curves remained separated after restriction to this cohort [Figure 2B], and IVIG use was associated with lower mortality after multivariable adjustment (HR, 0.15; 95%CI: 0.08-0.27; P < 0.001) [Table 3]. The unadjusted association was also evident among methylprednisolone-treated patients without MCS and among those receiving MCS [Figure 2C and D].

Overlap weighting substantially improved measured covariate balance in the methylprednisolone-treated cohort; every post-weighting absolute standardized mean difference was < 0.10 [Supplementary Table 4]. The overlap-weighted Cox model remained directionally consistent with the primary analysis (HR, 0.24; 95%CI: 0.07-0.87; P = 0.030) [Table 3].

The Aalen-Johansen cumulative incidence of ALI/AKI was lower among IVIG-treated patients in the overall cohort and in the methylprednisolone-treated strata with and without MCS [Figure 2E-G]. In the cause-specific Cox proportional-hazards model, IVIG use was associated with a lower hazard of ALI/AKI (HR, 0.43; 95%CI: 0.30-0.61; P < 0.001) [Table 3].

Among 79 ECMO-supported patients, 50 of 61 IVIG-treated patients (81.97%) and 6 of 18 non-IVIG-treated patients (33.33%) achieved successful weaning with survival to discharge. The cumulative incidence of successful ECMO weaning with survival to discharge was higher among IVIG-treated patients in the full ECMO cohort and after restriction to methylprednisolone-treated patients [Figure 2H and I]. In the Fine-Gray model, IVIG use was associated with a higher subdistribution rate of successful weaning with survival (sHR, 5.89; 95%CI: 2.96-11.73; P < 0.001). Cause-specific Cox, accelerated failure-time, and penalized logistic analyses yielded concordant estimates [Table 3].

Within the methylprednisolone- and MCS-treated subgroup, survival and ALI/AKI curves differed across dose categories [Figure 3A and B]; dose-stratified time-to-weaning curves are shown for the ECMO-supported subgroup [Figure 3C]. IVIG was initiated within 4 days of admission in 105 patients and after 4 days in 79 [Supplementary Table 5]. In descriptive analyses stratified by cumulative dose, initiation by day 4 was associated with more favorable survival and ALI/AKI curves than later initiation [Figure 3D-H].

Associations with mortality were directionally consistent using early-treatment thresholds of 3 days (HR, 0.16; 95%CI: 0.04-0.72), 4 days (HR, 0.15; 95%CI: 0.04-0.53), and 5 days (HR, 0.27; 95%CI: 0.10-0.73). In the day-4 landmark analysis, a cumulative dose of 10 to < 20 g was associated with lower subsequent mortality than < 10 g (HR, 0.26; 95%CI: 0.08-0.82; P = 0.022), whereas the estimate for ≥ 20 g was similar in direction but imprecise (HR, 0.26; 95%CI: 0.03-1.99; P = 0.193) [Supplementary Table 5]. These analyses were exploratory and did not identify a definitive dose or treatment window.

Among methylprednisolone-treated patients without MCS, admission LVEF was similar between groups, whereas the lowest in-hospital and discharge LVEF values were higher in IVIG-treated patients [Figure 4A]. In-hospital cTnI, NT-proBNP, and hsCRP levels were lower with IVIG [Figure 4B], and circulating cytokines were generally lower [Figure 4C]. The same pattern was observed among methylprednisolone-treated patients receiving MCS: admission LVEF was comparable, lowest and discharge LVEF values were higher with IVIG [Figure 4D]; cardiac injury and inflammatory biomarkers were lower [Figure 4E]; and all measured cytokines were lower in the IVIG group [Figure 4F].

Clinically indicated endomyocardial biopsy specimens were available for 11 patients (IVIG, n = 6; non-IVIG, n = 5), with a median admission-to-biopsy interval of 4.80 days [Supplementary Table 6]. IVIG-treated specimens had lower histological inflammatory scores and lower densities of myeloperoxidase-positive neutrophils, CD68-positive macrophages, and CD3-positive T cells; CD20-positive B-cell density did not differ significantly [Figure 5]. This selected tissue analysis was exploratory.

Across the evaluated strata of white blood cell count, alanine aminotransferase, aspartate aminotransferase, and creatinine, all mortality HR point estimates favored IVIG [Figure 6 and Supplementary Figure 1]. Adjusted HRs ranged from 0.17 to 0.30; these subgroup estimates were exploratory and were not interpreted as evidence of effect modification.