Introduction

Statins that competitively inhibit 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase are the most effective class of drugs currently available for lowering serum total cholesterol, low-density lipoprotein (LDL) cholesterol, apolipoprotein B, and triglyceride levels [1, 2]. Statins are also known to have cholesterol-independent pleiotropic effects, such as the improvement of endothelial function [3], reduction of oxidative stress and inflammation [4], regulation of angiogenesis [5], and reduction of thrombogenicity [6]. Sepsis is an infection-induced systemic inflammatory response syndrome. A delicate interaction exists between pathogens and host immune defense mechanisms during the course of infection [7]. Statins may be considered novel therapeutic agents that may aid the treatment and prevention of sepsis by targeting a number of inflammatory and immunomodulating cascades involved in the development of sepsis. Besides, statins counter the harmful effects of sepsis on the coagulation system by inhibiting the tissue factor and prothrombin fragment levels and increasing the expression of thrombomodulin [8–10].

Statin therapy was first described as a potentially effective treatment for sepsis in a murine model [11], but this potential role of statins remains controversial because of inconsistent relationships with outcomes observed in subsequent clinical studies [12]. Some previous studies documented improved outcomes and reduced inflammatory response in patients who developed sepsis while receiving statin therapy [13, 14], whereas others found no association or even increased mortality in patients with sepsis receiving statin therapy [15–17]. These conflicting data may be attributed to the examination of small samples over short follow-up periods, and insufficient consideration of the effects of various statin doses. The questions of which statin type and dose impact sepsis outcomes have remained unanswered [18]. To address these gaps in knowledge, we critically examined relevant evidence of the impact of statin potency, which is defined by statin type and dose, on clinical outcomes of patients with sepsis. This large-scale, population-based, propensity score-matching study included patients with sepsis who received prior statin therapy, identified using the Taiwan National Health Insurance Research Database (NHIRD). We compared the mortality benefits of low-potency and high-potency statins with those of no statin, and examined whether evidence supported an incremental benefit of high-potency over low-potency statins.

Methods

Data source

Taiwan’s National Health Insurance (NHI) program, launched in 1995, currently covers 99 % of the nation’s population of 23 million people. The NHI program offers comprehensive medical coverage, including that for outpatient, inpatient, emergency, dental, and traditional Chinese medicine services, as well as prescription drugs, to all Taiwanese residents. Each year, medical records from the NHI program were collected and transformed to registration files and original claim data for reimbursement by the Bureau of NHI. These databases (i.e., NHI enrollment files, claims data, and prescription drug registry) provide comprehensive utilization and enrollment information for all patients participating in the NHI program. The National Health Research Institute (NHRI) maintained and de-identified these data files such that all identifiable elements of personal data were scrambled, making it safe in terms of privacy to pursue research using the NHIRD. The identification of specific individuals was encrypted in a unique identification number that can link all information relating to each individual claim in the database. The confidentiality of the data abides by the regulations of the Bureau of NHI and the NHRI. Diseases were coded using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM).

Study design and data collection

The diagnosis of sepsis was defined by a primary discharge diagnosis of septicemia (ICD-9-CM Code 038.x) plus the prescription of antibiotics. Patients who were under 20 years of age and/or had a prior history of sepsis were excluded from the study. Therefore, the study cohort comprised patients with a first-time discharge diagnosis of sepsis between January 2000 and December 2010. If a patient was admitted for sepsis for multiple times in this period, only their first hospital admission was considered for inclusion in the analysis. Patients entered the cohort on the first day of hospitalization for sepsis and were followed until death, loss to follow-up, or 31 December 2011, whichever came first. The index date was defined as the first day of hospitalization for sepsis. Survival was measured from the index date to the date of last follow-up or death.

Exposure to statins and other drugs

Patients with the diagnosis of sepsis who had used statins continuously for 30 days before the index date were included in the statin users. The control group consisted of patients with the diagnosis of sepsis without using statins continuously for 30 days before the index date. Our analysis was based on the intention-to-treat principle [19], which means that statin users and non-users crossing over to the other treatment group or being lost to follow-up during the study period would still be included in the analysis as a part of their original treatment group. To clarify the effect of any dose–response relationship, we categorized statin users into two groups: those treated with high-potency statins (at least 10 mg rosuvastatin, at least 20 mg atorvastatin, or at least 40 mg simvastatin) and those treated with low-potency statins (all other statin treatments). Patients with sepsis who were treated with both high-potency and low-potency statins were categorized as high-potency statin users in analyses [20]. In accordance with the World Health Organization Anatomical Therapeutic Chemical Classification System [21], we identified patients’ use of medications that would potentially confound the association between statin use and sepsis outcomes, such as antiplatelet agents, warfarin, angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers (ARBs), beta blockers, calcium channel blockers, diuretics, nitrates, dipyridamole, steroids, estrogen, progesterone, non-steroidal anti-inflammatory drugs, proton pump inhibitors, and antihyperglycemic drugs.

Sociodemographic characteristics and comorbidities

Baseline sociodemographic characteristics, such as age, sex, income (less than NT$ 19,100, NT$ 19,100–41,999, and at least NT$ 42,000), and level of urbanization, were recorded. Urbanization levels were classified according to the system used in NHRI publications, which ranges from level 1 (most urbanized) to level 4 (least urbanized) [22]. The Charlson comorbidity index (CCI) was used to determine individuals’ overall systemic health, with each increase in the CCI score representing a stepwise increase in cumulative mortality [23]. We categorized the cause of sepsis according to infection site (respiratory tract, bacteremia, genitourinary tract, intra-abdominal, wound, central nervous system, device-related, endocarditis, and other). Other systemic diseases not included in the CCI, such as atrial fibrillation, valvular heart disease, hypertension, asthma, dyslipidemia, and drug abuse, were also taken into consideration.

Outcomes

The primary outcome was all-cause mortality within 1 year after hospitalization for sepsis. The analysis also included other adverse consequences of sepsis, including in-hospital death, intensive care unit (ICU) admission, shock events (use of inotropes and/or vasopressors following sepsis), and the use of mechanical ventilation during hospitalization.

Statistical analysis

In the current study, we performed 1:1 propensity score-matching analysis. The estimated propensity score was used to match each statin user to one non-user with similar propensity score based on nearest neighbor matching without replacement using calipers of width equal to 0.1 standard deviation of the logit of the propensity score. A statin user was selected for matching to a control subject with most similar propensity score in random order. If more than one control was eligible, one was selected at random [24]. Baseline characteristics of statin users and non-users were compared by standardized differences. The cumulative incidence of mortality in the statin user and statin non-user were calculated with the Kaplan–Meier method and were compared with the log-rank test. For matched data, Cox regression models with a conditional approach using stratification were used to calculate adjusted hazard ratios (aHRs) and 95 % confidence intervals (CIs) for associations between statin use and mortality [25]. We also used multivariate logistic regression to calculate adjusted odds ratios (aORs) and 95 % CIs for the associations between statin use and in-hospital death, ICU admission rate, shock events, and the use of mechanical ventilation. Finally, we used the likelihood ratio test to examine the interaction between the effect of statins on mortality and the following variables: age, sex, CCI score, diabetes mellitus, cerebrovascular disease, heart failure, chronic kidney disease, statin potency, continued statin use during hospitalization, and the year of index date. We further performed subgroup analysis based on the result of the interaction test.

Microsoft SQL Server 2012 (Microsoft Corporation, Redmond, WA, USA) was used for data linkage, processing, and sampling. Propensity scores were calculated using SAS software (version 12.0; SAS Institute, Inc., Cary, NC, USA). All other statistical analyses were conducted using STATA statistical software (version 12.0; StataCorp, TX, USA). Statistical significance was defined as p < 0.05.

Results

Characteristics of the study population

Figure 1 shows a flow chart of patient recruitment. After propensity score matching, we identified 27,792 pairs of statin users and non-users. A total of 18,007 (64.8 %) statin users were treated with low-potency statins and 9,785 (35.2 %) were treated with high-potency statins. The mean age of statin users was 69.1 (standard deviation, 11.8) years. A majority of patients were diagnosed with respiratory infection (41.6 %), followed by genitourinary tract infection (33.7 %) and intra-abdominal infection (11.8 %). The characteristics of the study population are shown in detail in Table 1.

Fig. 1
figure 1

Patient selection flow chart

Full size image
Table 1 Demographic and clinical characteristics of propensity score-matched patients with sepsis
Full size table

Risk of sepsis-related mortality

The overall mortality rate in statin users compared to non-users was 18.9 vs. 21.8 % at 1 month, 27.6 vs. 31.7 % at 3 months, 32.1 vs. 36.9 % at 6 months, and 36.6 vs. 42.4 % at 12 months. The risk of 1-year mortality was lower among statin users than among non-users (HR 0.83, 95 % CI 0.81–0.85, Table 2). In the analysis of different subtype of statins, we found that lovastatin (aHR 0.77, 95 % CI 0.70–0.84), simvastatin (aHR 0.79, 95 % CI 0.74–0.85), fluvastatin (aHR 0.81, 95 % CI 0.75–0.88), pravastatin (aHR 0.83, 95 % CI 0.75–0.92), rosuvastatin (aHR 0.83, 95 % CI 0.78–0.89), and atorvastatin (aHR 0.86, 95 % CI 0.82–0.90) were associated with a lower risk of 1-year mortality compared with statin non-users; however, the test of interaction was not significant (P interaction = 0.173). Compared with statin non-users, the 1-year risk of mortality was lower among low-potency (aHR 0.89, 95 % CI 0.85–0.93) and high-potency (aHR 0.80, 95 % CI 0.75–0.86) statin users (P interaction = 0.020; Fig. 3). In analyses using exposure to low-potency statins as a reference group, high-potency statins were associated with significantly lower risks of 1-year mortality (aHR 0.90, 95 % CI 0.86–0.94, Table 2). As shown in Fig. 2, Kaplan–Meier curves suggested that high-potency statin users had the best outcome, and statin non-users had the worst outcome.

Table 2 Incidence and risk of sepsis mortality among statin users and non-users
Full size table
Fig. 2
figure 2

Cumulative incidences for mortality of sepsis among statin users, statin non-users, low-potency statin users, and high-potency statin users

Full size image

In further analysis, we found that 10,084 patients (36.3 %) continued statin use during hospitalization among 27,792 statin users. Patients who continued statin use during hospitalization had a greater mortality risk reduction (aHR, 0.78, 95 % CI, 0.74–0.82) than those who discontinued statin use during hospitalization (aHR 0.86, 95 % CI 0.83–0.89, P interaction = 0.003, Fig. 3).

Fig. 3
figure 3

Forest plot of the effect of prior statin use on 1-year mortality in patients with sepsis by subgroups

Full size image

Risk of sepsis-related adverse consequences

Analyses of secondary outcomes yielded results similar to those for mortality. The rates of in-hospital death, ICU admission, shock events, and the use of mechanical ventilation in statin users compared to non-users were 25.1 vs. 28.1, 42.6 vs. 49.9, 36.2 vs. 37.4, and 30.0 vs. 31.1 %, respectively. Compared with non-users, statin users had significantly lower risks of in-hospital death (OR 0.86, 95 % CI 0.83–0.89), ICU admission (OR 0.95, 95 % CI 0.92–0.98), shock events (OR 0.95, 95 % CI 0.92–0.98), and the use of mechanical ventilation (OR 0.95, 95 % CI 0.92–0.99; Table 2). Compared with low-potency statin use, use of high-potency statins was associated with 15 % lower odds of in-hospital death (aOR 0.85, 95 % CI 0.80–0.90), 8 % lower odds of ICU admission (aOR 0.92, 95 % CI 0.87–0.97), 7 % lower odds of shock events (aOR 0.93, 95 % CI 0.88–0.98), and 14 % lower odds of the use of mechanical ventilation (aOR 0.86, 95 % CI 0.81–0.92; Table 2).

Discussion

In this nationwide propensity score-matching cohort study, we found that prior statin use for at least 30 days was associated with a 14 % reduction in risk-adjusted in-hospital death. The protective effects of statins seemed to persist even 1 year after discharge. High-potency statins were associated with a 20 % reduced risk of 1-year mortality after sepsis compared with non-users, and low-potency statins were associated with an 11 % reduced risk.

The need for further investigation of the prior or de novo use of statin therapy for sepsis is generally recognized, but these two applications have been regarded as distinct. Critical illness markedly impairs statin metabolism, resulting in decreased ubiquinone levels and mitochondrial and organ dysfunction [26, 27]. Thus, existing research findings raise concern that acute statin administration may negate the possible benefit of this drug or even cause harm. The present study focused on the prior statin therapy and reduction in sepsis-related mortality. We believe that prior statin therapy may be more beneficial in the early phases of the septic process, when inflammatory signals may be modified.

Previous studies have produced inconsistent results regarding the association between statin therapy and in-hospital death [16, 28, 29]. A hospital-based cohort study of 188 patients with sepsis found that statin use was associated with a decreased risk of in-hospital death [29]. In contrast, another study revealed that statin use was associated with worse outcomes and greater risk of in-hospital death in patients with sepsis receiving prolonged mechanical ventilation [16]. However, the ability to interpret the later study may be limited by insufficient consideration of clinical conditions. Recently, a published randomized controlled trial showed that the use of de novo statin therapy was ineffective in preventing mortality in ICU patients with ventilator-associated pneumonia [30]. The conclusion probably deserved to be extended to other types of nosocomial infection [30, 31]. According to previous studies [32, 33], about 5–10 % of patients admitted to the hospitals acquire one or more nosocomial infections. Thus, we suggested that most patients with hospital discharge diagnosis of sepsis in our study were admitted with sepsis rather than developed sepsis in the hospital. We found that prior statin use in patients with sepsis was associated with a 14 % lower risk of in-hospital death. Consistent with previous studies [34, 35], we also found that statin use was associated with lower risks of ICU admission and the use of mechanical ventilation after the diagnosis of sepsis; we also noted a reduced risk of shock events. Our study indirectly supports the concept that statin premedication may confer additional protection in community-acquired sepsis that is not related to atherosclerotic diseases.

Although Yang et al. [15] suggested that statin use was not associated with reduced 30-day mortality in Asian patients with sepsis, our study and a randomized controlled trial [14] documented a positive effect. Yang et al.’s [15] results may be explained by a lack of statistical power (only 104/454 patients took statins for at least 30 days before admission); the authors noted that more than 6,000–7,000 patients would be needed to reach significance when examining mortality in such a population [15, 36]. Thus, given that our sample comprised 27,792 patients with sepsis, our results may better reflect the real effects of statins in Asian patients.

In a recent multicenter randomized trial that included 250 critically ill patients, Kruger et al. [14] found that only the prior statin users rather than de novo users had a lower risk of 1-month mortality compared with placebo; however, there was no significant difference in 3-month mortality. In contrast to Kruger et al. [14], we found that statin use significantly reduced 3-month mortality in patients with sepsis. Although that trial had a randomized design, its results may be limited by the small number of patients and lack of examination of associated mortality as the primary outcome [14]. In the present study, we found that the effects of statins on survival benefit persisted for up to 1 year after sepsis treatment. Besides, similar to the findings of two previous studies [37, 38], our study found the survival benefit of continued stain use during hospitalization among prior statin users.

Recent studies have suggested that the anti-inflammatory effects of statins are associated with high-dose regimens [18, 39]. This study examines statin potency in the setting of sepsis-related mortality. The main advantage of using patients exposed to low-potency statins as a reference group was the expected elimination of a substantial amount of unmeasured confounding by indication. Our results add another dimension to the idea that high-potency statins are most beneficial. High-potency statins (at least 10 mg rosuvastatin, at least 20 mg atorvastatin, or at least 40 mg simvastatin) were associated with a 20 % reduced risk of 1-year mortality compared with non-users, whereas low-potency statins were associated with an 11 % reduction.

Importantly, our study provides evidence that high-potency statins offer a better survival benefit than low-potency statins do. However, several limitations of the current investigation are worthy of note. First, because we focused on prior statin use in patients with sepsis, we are unable to state conclusively whether acute statin administration also reduces sepsis-related mortality in hospitalized statin-naïve patients. Second, some patient information, such as premorbid activity level, tobacco use, and treatment adherence, and laboratory results were not available in the claims database. Thus, we could not extrapolate associations between sepsis-related outcome and white blood cell count, C-reactive protein level, or other infection profiles, as documented by other investigators. Additionally, data regarding the critical status of the patients (SOFA, APACHE II, or SAPS II scores calculation) were not recorded in the NHI database. Third, our study used an intention-to-treat analysis, and misclassification of some statin users, such as those who discontinued statins during the follow-up period, as non-statin users is more likely than the opposite situation. This bias could have contributed to underestimation of the protective effect of statins. Finally, our study was retrospective instead of prospective in nature. However, the use of propensity score-matching analysis allowed us to reduce or eliminate the effects of confounding when using observational data [40]. The strategy in the prescription of statins was based on doctors’ decisions, special considerations regarding allergy history or individual situation, and treatment adherence; however, such information was not available in our data set.

In conclusion, our results indicate that prior statin use is associated with a reduction in sepsis-related mortality that persists for at least 1 year. High-potency statins improved sepsis outcomes more effectively than low-potency statins did.