The study enrolled individuals with a confirmed diagnosis of APE with sepsis based on their discharge diagnosis. The diagnosis of sepsis is based on the Sepsis 3.0 criteria. Sepsis was defined as life-threatening organ dysfunction caused by a dysregulated host response to infection. For clinical operationalization, organ dysfunction was represented by an increase in the Sequential (Sepsis-related) Organ Failure Assessment (SOFA) score of 2 points or more (24). The inclusion criteria were: (1) APE had to be listed among the top five discharge diagnoses and had to be explicitly mentioned in the discharge diagnosis; and (2) the patients had to be 18 years of age or older. In cases where patients had multiple ICU admissions, only data from the initial admission were used. A comprehensive set of patients’ data was collected, including demographic details, vital signs, underlying conditions, laboratory results, clinical severity scores, and additional admission information. The diagnosis of APE was determined using the International Classification of Diseases, 9th and 10th editions.

The presence of statin medications in the “prescriptions” data from the MIMIC-IV database was used to assess the administration of statins. The statins included in the analyses were atorvastatin, simvastatin, rosuvastatin, lovastatin, pravastatin, and fluvastatin. The average daily dose was calculated to determine the dosage of statins. The classification of statin dosages was based on the potency of each statin, as indicated on a standard conversion chart (25).

The database contained variables previously reported to be cardiovascular risk factors and potential triggers for APE, as well as other variables (26–29). Personal demographic variables included age, sex, race, and body mass index (BMI). Health related variables included: respiratory rate, body temperature, the Saturation of Peripheral Oxygen ratio (SPO2), Sequential Organ Failure Assessment (SOFA) score, White Blood Cell (WBC) count, hemoglobin level, hematocrit, platelet count, and glucose level, and preexisting medical conditions (e.g., cardiovascular disease, kidney disease, rheumatic disease, liver disease, cancer, neurological disease, and chronic pulmonary disease).

The outcome variable was the probability of developing APE.

The study initially analyzed the baseline characteristics of the total sample and compared the characteristics of the two cohorts (Statin use and No statin use). Categorical data are summarized as frequency counts and percentages, whereas continuous data are presented as mean ± standard deviation or median (interquartile range), where appropriate. Analysis of variance or rank sum tests were performed to analyze differences in cohort outcomes for continuous variables. The Chi-square or Fisher’s exact tests were performed to analyze group (i.e., cohort) differences in outcomes for categorical variables.

We used a median replacement strategy to impute missing data on vital signs and laboratory parameters, as these variables contained missing data in 5% of the sample. Since the percentage of missing data for height and weight was low (ranging from 0.3–4%), no imputation was performed. We initially tested five multivariate logistic regression models to analyze the unique association between statins and APE, adjusting for different covariates. We performed some different statistical models to verify the results’ stability. In the final model, we adjusted the factors basing the following three rules (1 or 2 or 3). (1) We adjusted for variables, if it was added to this model, the matched odds ratio would change at least 10%. (2) For univariate analysis, we adjusted for variables, of which the p values were <0.1. (3) For multivariable analysis, variables were chosen on the basis of previous findings and clinical constraints. Supplementary analyses were conducted to examine subgroup and interaction analyses, controlling for relevant covariates. Propensity score matching (PSM) was conducted to improve the rigor of the study, using a 1:1 nearest neighbor matching algorithm with a caliper width of 0.1. Multivariate logistic regression models with robust variance estimators were employed to estimate the odds ratio (OR) for APE.

The statistical analyses were performed with STATA software (version 17.0), R packages (The R Foundation),¹ and Free Statistics software version 1.8 (30). Statistical significance was set to p < 0.05 (two-tailed).

Among the eligible patients, a total of 33,177 individuals met the sepsis criteria. After excluding cases of repeated ICU admissions and patients with an ICU stay of less than 24 h, the final cohort included 16,633 patients. Figure 1 presents a flowchart that illustrates the process of selecting study participants.

A total of 16,633 patients (57.3% male, mean age = 64.8 ± 16.2 years) were selected for inclusion in the study. The baseline characteristics of the study sample are presented in Table 1. Comparisons between the two groups indicated that the non-statin group was younger and had (a) a higher proportion of females, (b) higher SOFA scores, (c) a lower Charlson comorbidity index, (d) higher liver disorders rates and (e) significantly higher rates of APE, deep vein thrombosis (DVT), 30- and 90-day mortality. Longer ICU stays also were observed in the non-statin group. In the statins group, there were no significant liver or muscle-related side effects.

The univariate analyses found that the use of statins significantly reduced the rates of APE compared to no statin use (OR = 0.67, 95% confidence interval = 0.53–0.83, p < 0.001; Table 2).

The multivariate logistic regression analyses (Table 2) found the ORs for the benefit of using statins remained consistently significant across all five models (ORs ranged from 0.65 to 0.76, p < 0.05 for all models). Model 5, which controlled for all the covariates, found the use of statins had a significant 33% reduction in APE risk (OR = 0.67, p < 0.001), and these results were robust.

The multivariate logistic regression analyses for DVT of using statins remained consistently significant across all five models (ORs ranged from 0.36 to 0.53, p < 0.05 for all models). Model 5, which controlled for all the covariates, found the use of statins had a significant 47% reduction in DVT risk (OR = 0.53, p < 0.001; Supplementary Table S2).

The results remained consistent across the logistic regression models. After PSM was conducted on both groups, the sample consisted of 4,437 well-matched pairs, and there were no significant differences in key variables between the two matched groups (Supplementary Table S1). Among the 4,437 pairs in the propensity-matched pool, the risk of APE was significantly lower in patients who were prescribed statins [97 (2.2%) versus 141 (3.2%), p = 0.004]. The multivariate logistic regression model that was adjusted for all the covariates yielded an OR = 0.68 (p < 0.007) for APE (Table 2). Furthermore, when analyzing the net effect of the dosage of statins, both the standard dose of statin (OR = 0.72) and the high dose of statin (OR = 0.65) were associated with a reduced risk of APE (Table 3). Similarly, when analyzing the classification of statins, atorvastatin (OR = 0.62) had a protective effect, while simvastatin (OR = 0.87) and other statins (OR = 0.54) did not show a significant association with APE (Table 4). Subgroup analysis further supported the robustness and reliability of the observed statin-APE relationship. The protective effects of statins in these subgroup analyses were more pronounced in patients who also used oral anticoagulants than in patients who used non-oral anticoagulants. No other significant interaction was observed in the subgroup analyses (p for interaction >0.05) (Figure 2).

This current study builds upon previous promising findings about the use of statins for patients with sepsis. By utilizing a large-scale database, this study provides robust evidence supporting the favorable effects of statins in reducing the occurrence of APE in sepsis patients. The results of this study validate that administering statins is significantly associated with a substantial reduction in the likelihood of APE in sepsis patients. In addition, the subcategorization of statin usage, analysis of statin dosage, and PSM further strengthen the validity of these findings, by consistently demonstrating the protective effect of statins in reducing the risk of APE in patients with sepsis.

Extensive research has examined the impact of statins on APE (31–34), and the findings consistently demonstrate that statins effectively mitigate the occurrence of APE in populations at high risk, while also reducing mortality rates associated with APE. Consistent findings were observed in our study, revealing a significant association between the use of statins and a decreased risk of APE in sepsis patients (OR = 0.67). However, it is crucial to note that some studies have not replicated our findings. One study of the administration of statins within the initial year following a successful kidney transplant reported that statins did not reduce the likelihood of PE (35), and a study by Huerta et al. found no substantial protective effect of statins on APE or deep vein thrombosis in the context of current or past statin use (36). While a randomized clinical trial reported that the use of rosuvastatin substantially reduced the occurrence of symptomatic venous thromboembolism, its use did not result in a decline in the frequency of APE (37). However, that study had some limitations, including its sole focus on individuals without existing health issues, and its limited duration of observation. Furthermore, it did not examine the link between statin dosage and the probability of VTE. Another study demonstrated that the efficacy of statin therapy in preventing thrombus formation in cancer patients remains unclear (38). But, it is worth noting that the study only included participants with progressive tumor growth. Additionally, its small sample size and short duration of monitoring were study limitations. Therefore, additional clinical studies are needed to assess the efficacy of statins in the prevention and treatment of APE.

The precise mechanism through which statin use is associated with a reduced risk of APE in patients with sepsis remains unclear. However, apart from their lipid-lowering effect, statins also possess anti-inflammatory properties, leading to decreased levels of inflammatory markers in the blood and improved endothelial function. Moreover, statins exert antithrombotic effects and can regulate coagulation cascades through various mechanisms that are independent of alterations in cholesterol levels (39, 40). The protective effect of simvastatin on APE-induced pulmonary arterial pressure, hypoxemia, and inflammatory changes may be attributed to its modulation of the signaling pathway involving silent information regulator 2 (SIRT2) and nuclear factor-kappa B (NF-κB). Additionally, pretreatment with atorvastatin has been found to improve APE-induced pulmonary hypertension and increase 24-h survival rates by reducing the elevation of lung-activated matrix metalloprotein-9 following APE (41). The promising protective effects of simvastatin in patients with APE are linked to its modulation of the SIRT2/NF-κB signaling pathway. This is supported by its capacity to alleviate APE-induced pulmonary artery pressure, hypoxemia, and inflammatory changes, highlighting its potential therapeutic benefits (42).

Our study possesses several noteworthy strengths. First, it is worth noting, for example, though the effects of statin administration on APE have been extensively explored, there is a lack of conclusive evidence, specifically, patients with sepsis. Our findings shed light on the substantial reduction in APE risk associated with statin usage in sepsis patients. Second, the rationale for selecting statins lies in their widespread acceptance and ease of use within the medical community. Prior research has demonstrated the broad applicability of statins in the management and prevention of diverse conditions, including tumors, cardiovascular disease, and cerebrovascular disease (43–46). Third, we conducted several sensitivity analyses to ensure the robustness of our results.

These analyses are important for at least four reasons: (1) logistic regression analyses were adjusted using multiple models to control for potential confounding variables, and the stability of the results was confirmed using thorough model adjustments; (2) the analysis of the net effects of standard and high doses of statin, produced reliable findings, with the trend test indicating a more pronounced effect for high-dose administration; (3) the categorization of statin usage into non-use, atorvastatin, simvastatin, and others, revealed the protective effects of various statins against APE; and (4) the employment of PMS analysis yielded results consistent with those of the initial analyses.

We acknowledge several limitations of our study in line with previous observational studies. First, large amounts of missing data prevented us from conducting statistical analyses on lipid levels; therefore the optimal lipid value for APE in sepsis remains unknown. Second, the retrospective nature of our study and unmeasured confounders may have affected our findings. Furthermore, our analysis of serum markers of inflammation was limited to WBC. Interleukin-6, C-reactive protein, and procalcitonin were not measured due to a large proportion of missing values. A significant amount of data is also missing for risk stratification factors related to APE (echocardiography, electrocardiogram, CT - scan, BNP, and TNT). Third, our study may not be generalizable as it was conducted in a single institution in the United States. However, our substantial sample size and representative cohort lend support to our findings. Future prospective studies across multiple centers should help validate our findings. Fourth, we were not able to control for several potential confounding variables, such as smoking history, drinking history, hormone use, and other medical histories, which may have influenced the risk of APE in patients with sepsis. Additionally, the retrospective nature of our study prevents us from providing detailed information on the impact of statins on individual patients’ lipid levels, such as dosage and duration of use.