This study utilized data from the Medical Information Mart for Intensive Care IV (MIMIC-IV), a comprehensive single-center database containing electronic health records of patients admitted to the Beth Israel Deaconess Medical Center in Boston, Massachusetts, United States, from 2008 to 2019 (Johnson A. E. W. et al., 2023; Johnson et al., 2024). Access to the database was approved by the Institutional Review Boards of the Beth Israel Deaconess Medical Center and the Massachusetts Institute of Technology in Cambridge, Massachusetts, United States (record ID: 49780033). Owing to data anonymization, patient consent was not required to adhere to ethical standards and privacy regulations.

Participants were COPD patients aged 18 years and older who were admitted to the ICU under specific criteria: 1) a confirmed diagnosis of COPD, as indicated by a post-bronchodilator FEV1/FVC ratio of less than 0.70; 2) their first ICU admission; and 3) a minimum ICU stay of 48 h. Data collection included a wide range of parameters such as demographic details (age and sex), disease severity markers (Oxford Acute Severity of Illness Score (OASIS)), comorbidity burden (Charlson Comorbidity Index), organ support interventions (MV and continuous renal replacement therapy (CRRT)), baseline magnesium levels, and infection-related parameters (sepsis status and laboratory biomarkers). Laboratory indicators were initially measured within 24 h of ICU admission. For patients with multiple ICU admissions, only data from the first admission were considered to maintain data independence.

The main endpoint was the effect of magnesium sulfate treatment within 48 h of ICU admission on 28-day mortality. Secondary endpoints comprised the incidence of acute kidney injury (AKI), ICU mortality rate, ICU stay duration, and MV duration.

The covariate selection utilized the Boruta algorithm, least absolute shrinkage, and selection operator (LASSO) regression. LASSO regression employs L1 regularization to reduce certain feature coefficients to zero, aiding in both feature selection and parameter estimation. The Boruta algorithm functions as a robust wrapper method that statistically assesses original features against manually generated “shadow features” (randomized arrangements of the original data) using a Z-score test (Lin et al., 2025), effectively minimizing the false-positive rate. Ultimately, the variables “age, OASIS score, AKI, sepsis, MV, CRRT, blood urea nitrogen (BUN), creatinine” were incorporated into the regression model (Figure 1).

Continuous variables are reported as mean ± standard deviation or median interquartile range, depending on their distribution. Groups of variables were compared using Student's t-test for normally distributed data and the Wilcoxon rank-sum test for non-normally distributed data. Categorical variables are presented as counts and percentages and were analyzed using the chi-square test.

Three models were constructed to evaluate the impact of early magnesium sulfate administration on 28-day mortality: Model 1 remained unadjusted, Model 2 was adjusted for characteristic variables, and Model 3 included a propensity score methodology involving matching, covariate-adjusted scores, and inverse probability weighting. E-values were computed to assess the potential impact of unmeasured confounders on the relationship between early magnesium sulfate use and 28-day mortality (VanderWeele and Ding, 2017; Haneuse et al., 2019). Survival disparities were examined utilizing the Kaplan–Meier method and evaluated through time-series tests. Stratified analyses and interaction tests were carried out to confirm the consistency of associations across various subgroups, encompassing demographics, therapeutic interventions, sepsis, AKI, baseline magnesium levels, and OASIS scores. Data analysis was conducted using R software (version 4.4.2), with p-values below 0.05 considered statistically significant.

In this study, we analyzed a cohort of 3,651 patients with severe COPD. Within the first 48 h of ICU admission, 31.44% of the patients (1,148) received magnesium sulfate treatment, as detailed in Table 1. Our findings indicated that sepsis was diagnosed in 47.52% of the cohort (1,735 patients), and 71.87% (2,624 patients) developed AKI after admission. Both comorbidities occurred less frequently in the magnesium sulfate group than in the untreated group. This group also had significantly lower CRRT and invasive mechanical ventilation rates, as shown in Table 1.

Clinically, patients treated with magnesium sulfate demonstrated significantly lower mortality rates both in the ICU and at 28 days post-admission compared to those who did not receive the treatment, with p-values less than 0.05 for both endpoints (Table 2). Kaplan–Meier survival curves further emphasized a significantly higher 28-day survival rate among patients treated with magnesium sulfate, confirmed by a time-series test with a p-value of less than 0.05 (Figure 2). Additionally, the mean duration of MV was significantly shorter in the magnesium-treated group (p < 0.001). However, no significant difference was observed in the length of ICU stay between the two groups (p = 0.237).

Table 3 demonstrates a consistent link between lower 28-day mortality rates in all examined models for patients receiving magnesium sulfate treatment. Particularly, in the model utilizing propensity score inverse probability weighting, the use of magnesium sulfate was linked to a 24% reduction in 28-day mortality compared to non-users (hazard ratio [HR] 0.76 [95% confidence interval {CI} 0.60–0.95]). The calculated E-value of 1.96 (upper confidence limit 2.89) indicates that an unmeasured confounding factor would need to have a relative risk of at least 1.96 to negate the observed association.

The beneficial effects of magnesium sulfate on reducing 28-day mortality remained consistent across all subgroups analyzed Figure 3. Significant interactions were observed between magnesium sulfate administration and both BUN levels and mechanical ventilation use, with interaction p-values below 0.05, indicating statistical significance (Table 4). Specifically, patients with BUN levels under 24 mg/dL who received magnesium sulfate treatment exhibited a significantly reduced 28-day mortality rate (HR 0.48 [95% CI: 0.33–0.71]) compared to untreated patients. Similarly, a positive effect was noted in patients on mechanical ventilation who were treated with magnesium sulfate versus those who were not (HR = 0.54 [95% CI: 0.37–0.79]). These results highlight the substantial impact of magnesium sulfate in various clinical scenarios within ICU settings. Furthermore, while the relationship between magnesium sulfate use and 28-day mortality remained consistent across different baseline blood magnesium levels, no significant association was found in the subgroup with blood magnesium levels ≥2.4 mg/dL. Subgroup analyses of the neutrophil-to-lymphocyte ratio (NLR) and duration of mechanical ventilation indicated that patients treated with magnesium sulfate experienced lower 28-day mortality as NLR or duration of mechanical ventilation increased (Figure 4). Finally, we did not observe an increased risk of AKI associated with the early use of magnesium sulfate (Table 5).

Magnesium sulfate is extensively utilized in conditions like preeclampsia, eclampsia, and specific arrhythmias. Its role in critical care has expanded recently, driven by emerging evidence of broader benefits. For example, magnesium has been investigated as a treatment for atrial fibrillation (to aid in restoring sinus rhythm), kidney protection, and neuroprotection in ICU patients (Gu et al., 2023; Davey and Teubner, 2005; Saver et al., 2015). A retrospective cohort study utilizing the MIMIC-IV database discovered that magnesium sulfate usage in patients with sepsis was linked to significantly lower 28-day mortality, a decreased incidence of AKI, and the necessity for renal replacement therapy (Roussos and Koutsoukou, 2003). These benefits are ascribed to magnesium’s anti-inflammatory and antioxidant properties, which might mitigate organ damage seen in critical illness. Importantly, these advantages seem to depend on maintaining magnesium within an optimal range. Another analysis observed that excessive magnesium supplementation resulting in hypermagnesemia (serum magnesium >1.07 mmol/L) was associated with increased 28- and 90-day mortality (Li et al., 2024). This highlights that while magnesium sulfate is generally well tolerated, oversupplementation (particularly in patients with renal impairment) can lead to adverse effects such as hypotension and respiratory depression. Therefore, precise dosing and monitoring of serum magnesium levels are crucial when using magnesium in critical care.

The effectiveness of magnesium sulfate in acute COPD exacerbation has been explored in various trials and reviews with varying outcomes (Jahangir et al., 2022; Shivanthan and Rajapakse, 2014; Skorodin et al., 1995). A recent Cochrane systematic review and other meta-analyses on intravenous magnesium in COPD indicated some modest advantages: magnesium therapy enhanced lung function (e.g., increased FEV₁ and peak expiratory flow) and was linked to reduced odds of hospital admission (Ni et al., 2022). For instance, Jahangir et al. found that IV magnesium resulted in significantly improved FEV₁/PEF outcomes and approximately a 55% decrease in the risk of hospitalization in COPD exacerbation patients compared to placebo (Jahangir et al., 2022). Magnesium has long been recognized for its beneficial effects on severe asthma attacks by alleviating bronchospasm and enhancing FEV₁. However, it demonstrates limited efficacy in milder asthma exacerbations—a trend that appears to be somewhat mirrored in COPD research.

Despite these improvements in intermediate outcomes, previous randomized trials have not consistently shown improvements in the most critical endpoints of COPD exacerbation. Multiple RCTs, often small and single-center studies, have not demonstrated significant reductions in the need for ICU admission, intubation, or mortality with IV magnesium. For example, a trial involving 124 patients by Nouira et al. found no significant differences in intubation rates or 90-day mortality between the magnesium and control groups (Nouira et al., 2014). Similarly, an earlier trial by Skorodin et al. observed improved peak flow after magnesium administration but no significant impact on dyspnea scores or hospitalization rates (Skorodin et al., 1995). These inconsistent findings likely result from variations in study designs, such as differences in magnesium dosage (ranging from 1.2 g to 2.5 g), timing of administration, and patient populations, all of which can affect outcomes. Many RCTs have excluded the sickest COPD patients, such as those needing invasive mechanical ventilation or those with life-threatening respiratory failure. Therefore, previous trials may have lacked the power to detect survival benefits, particularly as they focused on moderate exacerbations treated in an emergency setting.

Our findings offer new insights in this context. In our analysis, encompassing a wide cohort of ICU patients with COPD exacerbation, the use of magnesium was linked to enhanced survival, particularly in the most severe cases necessitating mechanical ventilation. This differs from previous interventional trials that did not show a clear mortality benefit, underscoring that the life-saving potential of magnesium might only manifest in critical patient subgroups. In other words, magnesium sulfate could provide benefits in severe COPD exacerbations that were not evident in earlier studies focusing on less critical patients. By specifically investigating outcomes in mechanically ventilated individuals, our study indicates that magnesium might lower mortality rates where standard therapies often fall short. This new discovery warrants further exploration as it suggests that magnesium’s impact goes beyond minor improvements in lung function or hospital stay, potentially changing the trajectory of life-threatening COPD exacerbations.

The physiological effects of magnesium sulfate offer plausible explanations for its potential benefits in critically ill patients with COPD. Magnesium is a well-known smooth muscle relaxant in the airways, acting by antagonizing calcium influx into bronchial smooth muscle cells and inhibiting acetylcholine release, resulting in bronchodilation (Barbagallo et al., 2021; Knightly et al., 2017). During severe exacerbations, magnesium administration can alleviate bronchospasm, reducing airway resistance and dynamic hyperinflation, thus enhancing ventilation efficiency. This effect likely contributes to improved gas exchange and may lower the risk of barotrauma or ventilator-induced lung injury by decreasing the pressure needed to ventilate stiff, constricted lungs. Maintaining adequate magnesium levels has been linked to a shorter duration of mechanical ventilation in ICU patients, suggesting that magnesium-facilitated bronchodilation and lung mechanics can expedite respiratory recovery (Altun et al., 2019; Santosh Raju et al., 2023). Additionally, magnesium exhibits central nervous system and neuromuscular effects that could enhance patient-ventilator synchrony. By blocking N-methyl-D-aspartic acid receptors and reducing acetylcholine release at the neuromuscular junction, magnesium imparts mild sedative and muscle relaxant properties (Wang et al., 2011; Laires et al., 2004). These effects may help prevent episodes of dyssynchrony in mechanically ventilated patients. By promoting a calmer respiratory drive and relaxing respiratory muscles, magnesium might reduce the need for deep sedation or paralytics to achieve ventilator synchrony. Supporting this, a clinical study demonstrated that adding magnesium infusion to standard sedatives in critically ill patients significantly shortened the duration of mechanical ventilation and decreased sedative requirements (Altun et al., 2019). Improved ventilator synchrony and earlier weaning can lead to fewer ventilator-associated complications (such as ventilator-induced lung injury, nosocomial pneumonia, and prolonged sedation delirium) and ultimately enhance survival. Furthermore, the anti-inflammatory and antioxidant properties of magnesium are particularly relevant in COPD exacerbations, which are often characterized by intense airway and systemic inflammation. Magnesium can inhibit the release of pro-inflammatory cytokines such as TNF-α and IL-6, thereby attenuating the inflammatory cascade (Aryana et al., 2014; Cross et al., 2018). In patients with sepsis, this immunomodulation has been associated with improved organ function and reduced mortality; similar benefits may be observed in COPD exacerbations, where inflammation contributes to respiratory failure (Jain et al., 2024). Magnesium also acts as a cofactor for enzymes involved in scavenging reactive oxygen species, thereby reducing oxidative stress and tissue damage in the lungs (Liu and Dudley, 2020). By mitigating inflammation and oxidative injury, magnesium can help protect against multiorgan dysfunction during severe exacerbations and enhance the overall resilience of the patient.

In summary, the multifaceted actions of magnesium sulfate, including bronchodilation, modulation of the respiratory drive, and anti-inflammatory and antioxidant effects, provide a strong theoretical basis for the observed mortality reduction in patients with mechanically ventilated COPD. It is likely to improve acute respiratory physiology (facilitating ventilation and oxygenation) while addressing the systemic factors (inflammation and oxidative stress) that influence recovery. Careful monitoring is imperative, and even as we harness the therapeutic effects of magnesium, we must avoid hypermagnesemia (>2.4 mg/dL, or >1.0 mmol/L), which can cause hypotension, bradyarrhythmia, and respiratory depression. When used judiciously, magnesium sulfate has emerged as a promising adjunct in the management of severe COPD exacerbations, especially in patients requiring invasive ventilation, by improving both the pulmonary and systemic conditions that determine their outcomes. This hypothesis, supported by our data, adds a new dimension to the discussion on magnesium in COPD care and encourages future prospective trials targeting this high-risk population.

This study has several limitations. First, its retrospective single-center design inherently carries the risk of selection bias, limiting its generalizability to other institutions or healthcare systems. Despite applying multiple propensity-score-based methods and calculating E-values, our results remained susceptible to residual confounding from unmeasured variables. Specifically, the E-value analysis indicated that an unmeasured confounder would require a risk ratio of at least 1.96 with both treatment allocation and outcome to fully explain the observed association, suggesting that a substantial hidden bias would be necessary to nullify our findings. Second, detailed data on the exact timing of magnesium sulfate administration relative to key interventions, such as the initiation of mechanical ventilation, were not systematically recorded, precluding a more granular temporal analysis. Additionally, the administered dose of magnesium sulfate was not standardized; most patients received approximately 2 g intravenously; however, variations in dosing and duration prevented a meaningful dose–response evaluation. Third, we only extracted baseline serum magnesium levels measured within the first 24 h of ICU admission without information on intratherapy or post-therapy levels. Consequently, we could not assess dynamic changes in serum magnesium levels or the incidence of hypermagnesemia over the course of treatment. Fourth, potentially relevant concomitant medications such as corticosteroids, bronchodilators, antibiotics, and diuretics were not consistently available in the database and were thus not included in our propensity score models. Furthermore, certain disease severity indicators beyond the OASIS score, such as PaO₂/FiO₂ ratios or detailed ventilator parameters, were either incomplete or unavailable. Although our machine learning-based covariate selection and propensity-based adjustments were designed to mitigate such imbalances, the absence of these variables may introduce bias. Finally, the observational nature of the study precludes causal inferences. The findings should be interpreted as hypothesis-generating rather than definitive evidence of benefits.

Future research should aim to validate these results in multicenter prospective trials that incorporate standardized dosing protocols, serial magnesium concentration monitoring, and detailed temporal mapping of administration relative to critical interventions. Such studies should also explore subgroup effects, particularly in mechanically ventilated patients and those with varying baseline magnesium levels or renal functions, to refine patient selection and optimize the efficacy and safety profile of magnesium sulfate for ICU-managed COPD exacerbations.