1 Background

Statins, also known as 3-hydroxy-3 methylblycel coenzyme A (HMG-CoA) reductase inhibitors, play an important role in the prevention of coronary artery disease and ischemic stroke [1, 2]. In addition to their lipid-lowering effects, statins exert pleiotropic effects including reduction of inflammation, immunomodulation, antimicrobial effects, improved endothelial cell function and antithrombotic effects [3,4,5,6,7]. In view of the above mentioned effects of statins, intensivists have investigated whether statins are useful in critically ill patients. In patients with pneumonia, current statin use was associated with reduced mortality [8]. The use or previous use of statins may have reduced the mortality in septic patients in some observational studies [9,10,11]. Preliminary data from animal models have shown a protective effect on sepsisi induced lung injury [12]. The results of a meta-analysis support a similar conclusion [13]. However, some randomized controlled trials have shown the opposite [14,15,16,17,18]. In the Hydroxymethylglutaryl-CoA Reductase Inhibition with Simvastatin in Acute Lung Injury to Reduce Pulmonary Dysfunctin-2 Study (HARP-2), simvastatin compared with placebo was not associated with a reduced mortality in the overall population [18]. In the subset of patients with a hyperinflammatory phenotype, however, simvastatin was shown to reduce all cause mortality [19]. The differences in outcomes between the studies might be due to the characteristics of the different intensive care unit (ICU) populations. A recent retrospective study enrolled all ICU patients who had been prescribed statins and compared outcomes with a matched cohort. The results showed a beneficial association of statin use and 90 day mortality improvement [20]. However, the ICU population in this study may have been healthier or had more self-limiting disease processes than other ICU populations, such as patients with sepsis or acute lung injury/ARDS; this could have resulted in the relatively lower 90-day mortality observed in this study compared to that reported in other studies.

To avoid such patient selection bias and to, moreover, and also to investigate the mortality benefit of statin use in critically ill patients, we enrolled ventilated patients who had taken statins before or during ventilation as our study population. Ventilation is a key life-saving treatment measure for critically ill patients. However, mechanical ventilation is also associated with longer hospital stays, higher costs, and higher mortality rates [21]. Furthermore, ventilation can cause lung injury and other cardiovascular complications [22]. A recently published multicentre study included 3659 mechanically ventilated patients. This study found that the reason for intubation was stroke in 30.5% of the patients. During ventilation, cardiovascular events occurred in 42.6% of the 3659 patients [23]. These data suggest that statins may be able to improve outcomes in ventilated patients.

Whether statin use is associated with lower mortality in critically ill patients requiring ventilation in the ICU remains unclear. Therefore, we designed this observational study to research the potential beneficial effect of statin use among critically ill ventilated patients.

2 Methods

2.1 Study Design and Data Source

This is a retrospective observational study. We analyzed data from a large database: Medical Information Mart for Intensive Care (MIMIC-III). The MIMIC-III database is an openly available dataset developed by the MIT Laboratory for Computational Physiology, comprising deidentified health data associated with nearly 54,000 intensive care unit admissions [24]. The data in the MIMIC-III database consist of comprehensive clinical records of patients admitted to the ICUs of Beth Israel Deaconess Medical Center in Boston, MA, from June 1, 2001 to October 31, 2012. The requirement for institutional review board (IRB) approval from our institution was waived because MIMIC-III is a third-party anonymized publicly available database with pre-existing IRB approval.

2.2 Participants

Patients who underwent mechanical ventilation were selected from the MIMIC-III database. The inclusion criteria were as follows: (1) age ≥ 18; (2) patients who were put on mechanical ventilation; and (3) precise outcome data. The exclusion criteria were as follows: (1) age < 18; (2) no mechanical ventilation; (3) no precise outcome data; and (4) statin use after extubation. The ventilation data were extracted from the chartvcents table. Statin usage information was extracted from the prescription table. We selected those who had taken statins before or during ventilation as the statin cohort and those who underwent ventilation without statins as the nonstatin cohort. Those who took statin medicine after extubation were excluded from this study. Then, we performed a propensity matching. Each statin-exposed patient was matched with the closest corresponding nonexposed patient (that is, a patient who was not exposed to statins) at a 1:1 fixed ratio (nearest match cohort).

2.3 Outcomes

The primary outcome was 28-day and in-hospital all-cause mortality. The secondary outcome analyses also included ventilator-free days at 28 days and ICU-free days at 28 days, 60-day survival, 90-day survival and in-hospital survival. VFDs were defined as follows: (1) VFDs = 0 if the subject died within 28 days of mechanical ventilation; (2) VFDs = 28 − x if successfully liberated from ventilation x days after initiation; and (3) VFDs = 0 if the subject was mechanically ventilated for > 28 days. The definition of ICU-free days was similar to that of VFDs. Eight plasma biomarkers were extracted and studied to measure the host responses during the first 10 days after intubation. The HRs of each type of statin were studied by Cox models. The effects of statins in different subgroup populations were analyzed.

2.4 Statistical Methods

The doubly robust estimation method was applied to infer the sensitivity analysis of the primary outcome. “Doubly robust estimation combines a multivariate regression model with a propensity score model to estimate the association and causal effect of an exposure on an outcome” [25, 26]. Usually, the regression model or the propensity score model was applied individually to estimate a causal effect. When the two approaches were built in one estimation model, only one of the two models needs to be correctly specified to obtain an unbiased effect estimator; thus, the term doubly robust analysis. For propensity scores of statin use, a machine learning algorithm named the gradient boosted model (GBM) was employed to maximally correlate with the negative gradient of the predefined loss function. A regression tree was used, and a total of 35 covariates were used in the model so that covariate imbalance between the statin and nonstatin groups was minimized. Confounding covariates included (1) demographic characteristics (sex, age); (2) disease severity score (Simplified Acute Physiology Score); (3) history of coronary artery disease; (4) history of cerebral infarction; (5)primary diagnosis at admission; (6) other comorbidities at ICU admission (congestive heart failure, cardiac arrhythmias, valvular disease, pulmonary circulation, peripheral vascular, other neurological disease, chronic pulmonary, diabetes uncomplicated, diabetes complicated, hypertension, paralysis, hypothyroidism, renal failure, liver disease, peptic ulcer, lymphoma, metastatic cancer, solid tumor, rheumatoid arthritis, coagulopathy, obesity, weight loss, fluid electrolyte, blood loss anemia, deficiency anemia, alcohol abuse, drug abuse, psychoses and depression). Using the propensity score calculated by GBM, the statin cohort and nonstatin cohort were matched at a 1:1 ratio. The matching method was “nearest”, and the caliper value was 0.02. To evaluate the effectiveness of the propensity score model in balancing the two compared groups, the standardized mean difference (SMD) between the statin and nonstatin groups was calculated. Using the estimated propensity scores calculated by the GBM model as weights, an inverse probability weighting (IPW) model was used to generate a weighted cohort [26, 27]. An IPW weighted Cox regression was then built adjusting for the variables that remained unbalanced between the groups. The propensity score was also employed to build the 1:1 matching cohort of the nonstatin cohort.

The primary statistical method of comparison for the time-to-event end points was expressed by Kaplan–Meier curves and tested by the log-rank test. A Cox proportional hazards model was used to estimate the hazard ratio (HR) of 28 day mortality and its associated 95% confidence interval (CI). We included several variables in the model to adjust the 28-day survival, which was based on the top importance calculated by GBM (age, SAPSII score, sex, liver disease, diabetes, obesity, hypertension, renal failure, uncomplicated diabetes, complicated diabetes, coronary disease history, obesity, cerebral infarction history, congestive heart failure, cardiac arrhythmia, other neurological and chronic pulmonary disease). The APACHE III score was not recorded for every patient, so we could not extract all APACHE III scores. The SAPSII was chosen to represent the severity of illness.

Some studies have reported different potencies between statins; for example, simvastatin exerted better antibacterial effects than rosuvastatin, and the latter was found to have a more potent lipid-lowering capacity [28, 29]. Therefore, we analyzed the difference in 28-day survival among statin types in the Cox regression model. To test the efficiency of statins in patients with different profiles, we performed a subgroup analysis. Patient categorical data are presented as percentages, and continuous data are listed as the means with standard deviations (SDs). We used Student’s t tests for continuous variables and chi-square or Fisher’s exact tests for dichotomous variables. Rstudio1.2.1335 (RStudio, Inc., Boston, MA, USA) was used to perform the statistical analyses. The mean (standard deviation) was used for all continuous variables and with two decimal. And rates were compared using counts (percentages), with percentages retaining one decimal.

3 Results

3.1 Baseline Results

All 53,432 cases in the MIMIC-III database were screened. A total of 24,769 individuals who had undergone ventilation met the inclusion criteria. 3999 patients were selected for the statin cohort. A total of 17,065 patients who had not received statin treatment were included in the nonstatin cohort. We constructed a propensity score model by employing the 39 covariates with the GBM. The contributions of individual covariates to the final propensity score are illustrated in Fig. 1. The top covariates include diagnosis, age, history of hypertension, presence of CHF and peripheral vascular disease; such factors would commonly influence the decision regarding whether to prescribe statins. Based on the estimated propensity scores, IPW was applied to standardize the differences between the statin and nonstatin cohorts. As shown in Table 1, most of the covariates of the matched cohorts were balanced between the two cohorts with or without statins. There were 3999 patients who received statin treatment before or during the ventilation time. After matching, there were 3363 patients in each cohort (Table 1). They were similar in age, sex, SAPSII and 36 more variables. The characteristics of the patients are presented in Table 1. There were 2145 cases of atorvastatin, 200 pravastatin, 171 rosuvastatin, 1284 simvastatin and 199 cases who received other statins in the unmatched cohort. In the matched cohort, 1835 patients received atorvastatin, 162 patients received pravastatin, 136 patients received rosuvastatin, 1068 patients received simvastatin, and 162 patients received other statins. Here, patients who used two or more statins were classified as having other statins (Table 2).

Fig. 1
figure 1

Plot of the relative influence factor of covariates. The relative influence factor measures how discriminative the 35 covariates of the propensity score model are when predicting the likelihood of statin prescription

Full size image
Table 1 Demographic data
Full size table
Table 2 Details of statin use
Full size table

3.2 The Primary Outcome and Doubly Robust Analysis

In the overall cohort, the crude 28-day risk of all-cause mortality in statins vs nonstatins was 7753 (44.4%) vs. 1365 (34.1%), p < 0.01. The Kaplan–Meier analysis showed that statin users had a significantly improved 28-day survival (p < 0.01, Fig. 2A). The corresponding unadjusted hazard ratios (HRs) and 95% CIs were 0.85 (0.80–0.90) at 28 days.

Fig. 2
figure 2

Kaplan–Meier curves for all-cause mortality in statin recipients vs. nonrecipients in the unmatched cohorts and propensity score-matched cohorts A Statin use was associated with improved 28 day survival in the unmatched cohort (HR 0.85 95% CI 0.80–0.90); B Statin use was associated with improved 28 day survival in the matched cohort (HR 0.79 95% CI 0.73–0.85)

Full size image

In the matched cohort. The 28-day mortality was 1599 (47.5%) vs. 1230 (36.6%) in statin users vs. nonstatin patients. The Kaplan–Meier analysis showed that statin users had a significantly improved 28-day survival (p < 0.01, Fig. 2B). The corresponding unadjusted hazard ratios (HRs) and 95% CIs were 0.79 (0.73–0.85) at 28 days.

In the multivariate Cox model, the use of statins was associated with a beneficial effect on 28-day survival in the unmatched cohort (HR 0.73 95% CI 0.68–0.77 Table 3). Under the doubly robust estimation framework, univariate and multivariate regression models were developed to adjust for these unbalanced covariates in the weighted cohort (Table 3). The univariate model showed that the HR of statins was 0.85 (95% CI 0.80–0.90, p < 0.01, Table 3), whereas in the multivariate weighted model, it was 0.85 (95% CI 0.82–0.88, p < 0.01, Table 3).

Table 3 Five different models of primary outcome analysis
Full size table

3.3 Secondary Outcomes and Plasma Biomarkers

The secondary outcomes included VFDs and IFDs. After matching, statin use was associated with more VFDs (14.93 ± 13.11 vs 12.06 ± 13.26) and more IFDs (13.41 ± 12.14 vs. 10.86 ± 12.19) (Table 4). We analyzed the effect of statins on survival in patients with different lengths of stay in the matched cohort. Statin use was associated with improved 60-day (HR 0.81, 95% CI 0.75–0.86, p < 0.001), 90 day (HR 0.81, 95% CI 0.76–0.87, p < 0.001) and in-hospital survival (HR 0.81, 95% CI 0.76–0.87, p < 0.001) in the matched cohort (Table 4).

Table 4 The secondary outcomes
Full size table

The results of the biochemical tests all followed a normal distribution and we used box plots to represent these continuous variables and t-tests to calculate the differences between the means. We found no large differences in the matched cohort. Additionally, we found that statin use was not associated with higher aspartate aminotransferase (AST) or alanine aminotransferase (ALT) levels after matching (Fig. 3). The WBC and neutrophil counts were significantly lower in the statin cohorts (Fig. 3). However, creatinine was slightly higher in the statin group (Fig. 3).

Fig. 3
figure 3

The plasma biomarkers during the first ten ventilation days. Day 1 was the first day of ventilation; *p < 0.05; **p < 0.01.T-tests are used to compare means between two groups

Full size image

3.4 Sensitivity and Subgroup Analysis

We further analyzed the different statins and their effect on patients' 28-day survival. All the different kinds of statins were found to be associated with improved 28-day survival of the patients. We compared users of different types of statins to nonusers by the Cox regression model. In the Cox model, all statins improved the survival of ventilated patients. Other statin groiup seemed to be associated with the best outcome (HR 0.48 95% CI 0.35–0.65), followed by rosuvastatin (HR 0.59 95% CI 0.41–0.85), pravastatin (HR 0.65 95% CI 0.49- 0.87) and simvastatin (HR 0.72 95% CI 0.64–0.81). The least-effective statin was atorvastatin, which was associated with decreased 28-day mortality (HR 0.84 95% CI 0.77–0.92). After adjusting for confounders, all statins still resulted in significant improvement in survival compared to nonuse based on this survival function (Fig. 4A).

Fig. 4
figure 4

Plots of subtype analysis and subgroup analysis. A The different types of statin use and survival expressed by Kaplan–Meier curves. B The forest plot of statins in the 28-day survival subgroup analyses

Full size image

In the subgroup analysis, we found that statins had a significant interaction effect in subgroups of myocardial infaction. Statin use showed no significant interaction effect in inflammation-related groups, such as the sepsis, or pneumonia subgroups (Fig. 4B).

4 Discussion

This study revealed that statin use was associated with improved 28-day survival and in-hospital survival. All types of statins were associated with reduced reduced mortality. This evidence might indicate a protective effect of statins in patients on mechanical ventilation. The use of statins may also be associated with longer ventilator-free days and longer ICU-free days. The protective effect of statins in critically ill patients may only be applicable to a specific population [19]. We selected ventilated patients for the study for three reasons: (1) ventilated patients represent patients with severe illness and high mortality. Selective bias was avoided; (2) one-third of the ventilated patients would have been intubated for stroke [23]; and (3) ventilated patients had a 40% chance of having a cardiovascular event [23]. Therefore, ventilated patients are a more appropriate choice as a study population for statin exposure factors.

In addition to lowering cholesterol, statins exert pleiotropic effects [3,4,5,6] such as anti-inflammatory, antioxidant, and immunomodulatory effects, particularly in the context of lung disease [30]. Statins may reduce COPD exacerbation [31]. Some observational studies have also suggested that statins may be effective in patients with acute lung injury or acute respiratory distress syndrome [32, 33]. These clinical effects may be mediated by a reduction in pulmonary and systemic inflammation. Simvastatin reduced bronchoalveolar lavage IL-8 by 2.5-fold (P = 0.04) [34]. Statins also showed a protective effect against sepsis. Compared with non-users, simvastatin (HR, 0.72; 95% CI, 0.58–0.90) and atorvastatin (HR, 0.78; 95% CI, 0.68–0.90) users had improved 30 day survival [10]. The current study investigated the anti-inflammatory effect of statins in a cohort of patients with mechanical ventilation, which has not been previously reported. A recent study showed that infections in older adults were associated with prolonged, impaired neutrophil migration. Simvastatin improves neutrophil migration in vivo in healthy individuals and in vitro in milder infectious events but not in severe sepsis, supporting its potential utility as an early intervention in pulmonary infections [35, 36]. Lung injury remains one of the major complications of mechanical ventilation in the ICU. This injury may result from an altered host immune response after mechanical stretch [37]. The use of statin therapy to protect patients with lung injury could therefore be a reasonable strategy, as these drugs may attenuate the host inflammatory response to infection [13], especially within the lungs [3].

Animal models may have further explained the protective effect of statins against lung injury. Statins increase glucocorticoid receptor expression in alveolar macrophages and downregulate NF-κB activation, which is associated with an increased number of alveolar macrophages [12]. Two other animal models of mechanical ventilation-induced lung injury also support these findings [38, 39]. The protective effect may be due to the anti-inflammatory effect of statins. Prior statin use was associated with a lower baseline IL-6 levels, and continued atorvastatin treatment in this cohort was associated with improved survival [40]. Statins have been shown to reduce vascular leakage and inflammation in animal models of lung injury [41]. Statins may also attenuate lung injury by downregulating the expression of inflammatory cytokines [42, 43]. Our previous study demonstrated the lung-protective effect of statins caused by reducing in inflammatory cell infiltration [44]. In addition, statins may have direct antibacterial effects and modulate bacterial virulence [45,46,47]. Sarah et al. showed that prior exposure to physiological nanomolar serum concentrations of simvastatin confers significant cellular resistance to the cytotoxicity of pneumolysin, demonstrating how statins contribute to the reduced pathology observed in the context of pneumonia and other bacterial infections [48].

The results of the subgroup analysis showed no interaction between statin treatment in the pneumonia subgroup and the sepsis subgroup. This is consistent with other negative findings obtained in pneumonia and sepsis [14,15,16,17,18]. In a randomized controlled trial (RCT) with more than one year of follow-up, there was no significant difference in cumulative survival between the rosuvastatin and placebo groups (58 vs 61%; p = 0.377) [49]. Simvastatin therapy was not significantly associated with the difference in 28-day mortality (22.0 and 26.8%; P = 0.23) among patients with ARDS [18]. The conflicting results may be due to the different study populations. These RCTs have focused on patients with severe conditions such as severe ARDS and sepsis. Similarly, the subgroup analyses in this study showed that statins were not effective in the context of pneumonia, sepsis or respiratory failure. Sapey stated that statins may improve neutrophil migration and may have a protective effect in milder infectious events but not in severe sepsis or ARDS [35]. The reason for this controversial evidence may be that statins may have immunomodulatory effect only in milder diseases instead of in severe inflammatory diseases such as ARDS. However, new RCTs investigating the effects in milder infectious diseases such as ventilator-associated lung injury may be needed to prove this hypothesis. About 30–40% of the cases included were patients with myocardial infarction and cerebrovascular disease. Statins are known to improve outcomes in patients with cardiovascular and cerebrovascular disease; therefore, in the subgroup analysis, we focused on people with myocardial infarction and cerebral infarction. Our findings showed a significant association between statin and reduced mortality in patients on ventilators, regardless of the presence of myocardial infarction or cerebral infarction. There was a significant interaction effect of statin use on and between subgroups. This was because statin use resulted in a greater benefit for patients with myocardial infarction.

The current study has several limitations. The main limitation of this study is that its observational nature without randomisation precludes a definitive conclusion on the benefit of statins. However, a randomised controlled trial of the effect of long-term statin treatment on the outcome of patients on mechanical ventilation would require a large number of participants. For the time being, observational data may remain the best available evidence to investigate the effect of statins on patients on mechanical ventilation. Second, because of the retrospective design of this study, patient selection bias may be unavoidable. Third, the lack of data on potential confounders was a limitation that could not be overcome. Therefore, our results should be interpreted with caution. Regarding the host response, the assessment of inflammatory cytokines may provide different insights. The relationship of statin dose and duration with survival was not analysed here because we included different types of statins and the doses of the different statins were not comparable. Finally, we included patients on mechanical ventilation with different diagnoses. Even with subgroup analyses, we cannot conclude that statins are specifically effective in a particular population. Further studies in different populations or RCTs are needed to validate our findings.

5 Conclusions

Our study suggests that statin use may be associated with reduced mortality in ventilated patients. In addition, statin use may be associated with reduced ventilator use and length of stay in the ICU.