Statins for sepsis: a critical and updated review

Authors

  • P. Kopterides,

    1.  2nd Critical Care Department, ‘Attiko’ University Hospital
    2.  Alfa Institute of Biomedical Sciences (AIBS)
    Search for more papers by this author
  • M. E. Falagas

    1.  Alfa Institute of Biomedical Sciences (AIBS)
    2.  Department of Medicine, ‘Henry Dunant’ Hospital, Athens, Greece
    3.  Department of Medicine, Tufts University School of Medicine, Boston, MA, USA
    Search for more papers by this author

Corresponding author and reprint requests: M. E. Falagas, Alfa Institute of Biomedical Sciences (AIBS), 9 Neapoleos Street, 151 23 Marousi, Greece
E-mail: m.falagas@aibs.gr

Abstract

There is increasing discussion of a potential role for statins in the management of sepsis. A search of PubMed, Embase, Scopus and the Cochrane Library databases was performed by combining the terms ‘statins’, ‘infection’, ‘sepsis’, ‘bacteraemia’, ‘pneumonia’, and ‘ICU infections’. A total of 22 studies were retrieved, which included 177 260 people and compared clinical outcomes between 51 193 statin users and 126 067 non-statin users. Nineteen were cohort studies (seven prospective and 12 retrospective), two were retrospective case–control studies, and one was a randomized controlled study. Nine studies examined the use of statins in sepsis, four in community-acquired pneumonia (CAP), three in bacteraemia, and three in post-operative patients. Mortality data were presented in 15 studies; in ten, mortality was lower among statin users (three of six sepsis studies, five of six CAP studies, and two of three bacteraemia studies). In four studies, there was no difference in mortality (two of six sepsis studies, one of six CAP studies, and one of three bacteraemia studies) and in one study there was increased mortality among septic intensive-care unit patients receiving statins. Five of the nine studies that examined the risk of developing sepsis/infection as a primary outcome (six of nine sepsis studies and all studies in the postoperative setting) found a decreased risk among statin users, whereas the remaining studies found no difference. Irrespective of their design (matched vs. non-matched), the majority of the studies suggested that statins have a beneficial effect on the outcome of infection; however, their observational design does not allow us to draw firm conclusions. The clinical benefit of statin therapy in sepsis remains to be determined by ongoing randomized controlled trials.

Introduction

Sepsis is a common, expensive and frequently fatal clinical problem. Previous studies have estimated that more than 750 000 episodes of severe sepsis (300 cases per 100 000 population) occurred in the USA in 1995 and approximately 215 000 of the patients died [1]. At that time, the average cost per case was $22 100, and this resulted in an economic burden of nearly $17 billion annually in the USA alone [1]. Disturbingly, the incidence of sepsis and the number of sepsis-related deaths appeared to be increasing by 8.7% each year from 1979 through 2000 [2], and there is no sign of a reversal of this trend.

Sepsis is a highly complex inflammatory syndrome in which multiple cellular and humoral pathways are involved. Merely interfering with one of these pathways may be insufficient to arrest the whole inflammatory process, and this may partly explain why most of the adjunctive therapies developed for severe sepsis have shown disappointing results in rigorous clinical trials [3]. In fact, treatment with activated protein C is the only adjunctive therapy shown to be effective [4]; the initial enthusiasm about the role of steroids [5] has been tempered by the negative findings of the CORTICUS study [6].

Statins are lipid-lowering drugs that inhibit 3-hydroxy-3-methylglutaryl coenzyme A reductase and reduce total cholesterol, low-density lipid cholesterol, apolipoprotein B and triglyceride levels. These drugs have an excellent track record in reducing cardiovascular mortality among patients with and without coronary artery disease [7]. In the last decade, it was shown that statins possess a multitude of pleiotropic anti-inflammatory/immunomodulatory effects that are independent of their lipid-lowering effect; these immunomodulatory effects appear to be mediated through the blockade of mevalonate synthesis. Mevalonate is the precursor of multiple isoprenoid products that, in turn, facilitate the intracellular trafficking of cellular membrane proteins with signalling functions. For example, the inactivation of small GTP-binding proteins, which work as molecular on/off switches for cytoplasmic kinases—nuclear factor-κB, the mitogen-activated protein kinase family, and protein kinase B—leads to attenuation of this proinflammatory signalling and moderation of the maladaptive inflammatory response. Although a detailed discussion of the mechanisms by which statins may influence sepsis-associated inflammation is beyond the scope of this review, it should be noted that statins lower the proinflammatory tendency of leukocytes and monocytes, enhance T-helper cell function, directly suppress major histocompatibility class II antigen expression by macrophages, reduce the release of cytokines, chemokines and acute-phase reactants, impair the expression of adhesion molecules such as P-selectin, limit endothelial cell activation and improve endothelial function by reducing the inducible nitric oxide synthase/endothelial constitutive nitric oxide synthase ratio, modulate coagulation by blunting tissue factor and plasminogen activator inhibitor-1 expression and enhancing protein C function, display antioxidant and antiapoptotic properties, and activate haem oxygenase, an inducible, heat shock cytoprotective protein [8].

The pleiotropic actions of statins provided the rationale for investigating their role in sepsis, not only in vitro, but also in animal and clinical studies. The intense interest in their use for the management of septic patients has led to a series of publications with contradictory results that leave the clinicians hopeful but confused. Thus, we sought to critically examine the relevant evidence from studies regarding the role of statin use in the management of sepsis.

Literature search

We updated and extended our previous electronic search [9] through PubMed, Embase, Scopus and the Cochrane Library databases using the following key terms: ‘statins’, ‘infection’, ’sepsis’, ‘bacteraemia’, ‘pneumonia’, and ‘ICU infections’. We focused on peer-reviewed, full-text, observational cohort or case–control studies and randomized controlled trials that compared infection-related outcomes (i.e. hospital mortality) among patients receiving or not receiving statins. We did not, a priori, exclude any specific subpopulation of patients from our dataset. For the purpose of this review, we focused on bacterial infections, and excluded studies investigating the effect of statins on viral (i.e. influenza and human immunodeficiency virus infection), fungal and protozoan infections. We also excluded studies that were experimental or laboratory-based. On the other hand, we set no time or language limit on our search, and scanned all references from identified articles for additional relevant information.

We extracted and tabulated data concerning study design, patient settings, type of infection (sepsis, bacteraemia, pneumonia, and postoperative infections), number of included patients, and primary and secondary clinical outcomes. We also focused on the methodological aspects of the included studies, especially the inclusion of propensity scoring as an attempt to control for confounding bias.

Finally, we searched the online registry of randomized controlled trials (RCTs) of the US National Institutes of Health (http://www.clinicaltrials.gov) and the Current Controlled Trials website (http://www.controlled-trials.com) for relevant studies. The end date of the review was 15 December 2008.

Available Relevant Studies

We retrieved 22 studies [10–31] regarding the use of statins in patients at risk of infection that included a total number of 177 260 individuals (51 193 statin users and 126 067 non-statin users). The study cohort included 72 842 admitted patients and 104 418 individuals who were followed in the outpatient setting.

Before discussing these studies, it is necessary to briefly present five additional studies [32–36] that we excluded for various reasons from further analysis; some of them [32–34] have been included in similar reviews of the subject. Frost et al. [32] conducted a matched-cohort study and a separate case–control study (including 397 influenza deaths), and found a significantly reduced risk of unspecified pneumonia/influenza death among moderate-dose statin users. However, the authors did not discriminate between viral and bacterial pneumonia, based their diagnosis on assigned computerized codes, and did not verify compliance with the prescribed medications. In a retrospective cohort study, Schmidt et al. [33] showed that patients who developed multiple organ dysfunction syndrome under statin treatment had lower 28-day mortality rates as well as lower hospital mortality rates. Unfortunately, they did not specify whether infection was the underlying cause of multiple organ dysfunction syndrome. Ishida et al. [34], by implementing a novel idea, designed a population-based analysis using data from marketing surveys and government reports on mortality; they considered the statin use as expressed by statin sales per capita in the aged population and the mortality from major causes of death among 47 prefectures in Japan, and found a correlation between statin sales and decrease in mortality (p <0.05).

Donnino et al. [35] performed a secondary analysis of a prospective, observational study conducted at an urban, academic hospital emergency department, and found that, among 2036 patients with suspected infection, patients who received statins had a statistically significant lower mortality rate than those who did not (1.9% vs. 4.4%; p 0.02). The findings of this study have been presented only in an abstract. Finally, Ali and Buth [36] studied 5469 patients (3555 were receiving statins and 1914 were not) who underwent coronary artery bypass grafting (CABG), valve or combined CABG/valve surgery, and found that the unadjusted rate of sepsis and/or deep sternal wound infections was lower among statin users (2.3% vs. 4.1%; p 0.0002); however, when propensity score analysis was used to match the two subgroups of patients (1443 receiving statins and 1443 not receiving statins), statin use was not associated with a reduction in sepsis and/or deep sternal wound infections (OR 0.7; 95% CI  0.5–1.1; p 0.1), or mortality rate (36.3% in both groups; p 1.0). These findings have also been presented in abstract form only (the manuscript is in preparation; I. Ali, personal communication).

Nineteen of the 22 included studies were cohort studies (seven prospective [14,15,18,21,23,26,29] and 12 retrospective [10,11,13,16,17,19,20,25,27,28,30,31]) and two were retrospective case–control studies [22,24]. We identified only one randomized controlled trial [12]. The evaluable studies were divided into four groups according to the type of infection and tabulated accordingly: (i) studies regarding the use of statins in patients with sepsis or at risk of sepsis (Table 1); (ii) studies including patients with community-acquired pneumonia (CAP) (Table 2); (iii) studies referring to patients with bacteraemic infections (Table 3); and (iv) studies that focused on postoperative infections (Table 4).

Table 1.   Clinical evidence regarding the use of statins in sepsis
StudyStudy design Patients’ settingsStudy groupsPrimary outcomeSecondary outcomes
[10]Retrospective national cohort study using the Department of VA national patient care and pharmacy databases3018 patients hospitalized with sepsis in 2000, who had at least 1 year of previous VA outpatient care, and at least one active and filled VA prescription within 90 days of admissionSUs: 480
NSUs: 2538
Statin use was significantly associated with decreased 30-day mortality (OR 0.48; 95% CI  0.36–0.64)Angiotensin II receptor blocker use was also significantly associated with decreased 30-day mortality (OR 0.42; 95% CI  0.24–0.76)
[11]Retrospective cohort study454 patients with sepsis at National Taiwan University HospitalSU: 104
NSU: 350
No significant difference in 30-day sepsis-related mortality between groups (19.2% in SU vs. 18.9% in NSU; p 0.952)Statin treatment was not associated with decreased 30-day mortality (RR  0.95; 95% CI  0.53–1.68; p 0.853)
[12]Randomized controlled trial80 patients with subarachnoid aneurysmal haemorrhage randomized to receive pravastatin or placebo for up to 14 daysSUs: 40
NSUs: 40
Development of sepsis was not different (27.5% in SUs vs. 15.0% in NSUs; p 0.176)Pravastatin reduced sepsis-related mortality (6.25% vs. 71.43%; log-rank test, p 0.001)
[13]Retrospective cohort study53 patients admitted with sepsis to a university-associated teaching hospitalSUs: 16
NSUs: 37
Rate of severe sepsis was lower in SUs (56.0% vs. 86.0%; p <0.02)In-hospital mortality rates were similar (38.0% vs. 49.0%; p 0.33), whereas rate of cardiovascular dysfunction was lower in SUs (38.0% vs. 73.0%; p <0.02)
[14]National prospective cohort study1041 incident dialysis patients recruited from 81 US not-for-profit outpatient dialysis clinicsSUs: 143
NSUs: 898
Sepsis-related hospitalizations were significantly lower in SUs (crude incidence rate, 41/1000 patient-years) than in NSUs (crude incidence rate, 110/1000 patient-years) (p <0.001). With adjustment for demographics and dialysis modality, SUs were substantially less likely to be subsequently hospitalized for sepsis (incidence rate ratio 0.41; 95% CI  0.25–0.68) 
 Propensity-matched subcohort was studied SUs: 107
NSUs: 107
In the propensity-matched subcohort, statin use was even more protective (incidence rate ratio, 0.24; 95% CI  0.11–0.49) 
[15]Prospective, observational, population-based study
Propensity score was used
11 362 patients with atherosclerotic diseases identified and followed for up to 3 yearsSUs: 5698
NSUs: 5664
Infection-related mortality was significantly lower in SUs than in NSUs (0.9% vs. 4.1%; RR  0.22; 95% CI  0.17–0.28)
Stepwise Cox proportional hazard survival analysis including the propensity score for receiving statins revealed that the protective effect of statin use adjusted for all known potential confounders remained highly significant (HR 0.37; 95% CI  0.27–0.52)
 
  1. HR, hazard ratio; ICU, intensive-care unit; NSU, non-statin user; RR, relative risk; SU, statin user; VA, Veterans Affairs.

[16]Retrospective cohort study438 patients at high risk of ICU-acquired infections, i.e. those receiving mechanical ventilation for >96 hSUs: 38
NSUs: 400
Hospital mortality was significantly higher in SUs (61.0% vs. 42.0%), even after adjustment for APACHE II predicted risk (p 0.03)The ICU-acquired infection rate in SUs was non-significantly lower (29.0% vs. 38.0%, p 0.3) and delayed (median 12 vs. 10 days, p 0.6)
[17]Retrospective, population-based, cohort study
Propensity score was used
69 168 patients hospitalized for acute coronary syndrome, ischaemic stroke, or arterial revascularizationSUs: 34 584
NSUs: 34 584
71.2 events of sepsis per 10 000 person-years in SUs vs. 88.0 per 10 000 person-years in NSUs (p 0.0003)The protective association between statins and sepsis persisted in high-risk subgroups, including patients with diabetes mellitus, chronic renal failure, or a history of infections
[18]Prospective observational cohort study361 patients enrolled with pneumonia, cellulitis, or urinary tract infectionSUs: 82
NSUs: 279
Severe sepsis developed in 2.4% of SUs vs. 19.0% of NSUs (RR  0.13; 95% CI  0.03–0.52; p <0.001)The ICU admission rate was 3.7% in SUs vs. 12.2% in NSUs (p 0.025), reflecting an RR of ICU admission of 0.30 (95% CI  0.1–0.95)
Table 2.   Clinical evidence regarding the use of statins in community-acquired pneumonia
StudyStudy designPatients’ settingsStudy groupsPrimary outcomeSecondary outcomes
[19]Retrospective population-based cohort study29 900 adults hospitalized with CAP in northern DenmarkSUs: 1372
NSUs: 28 528
Mortality among SUs was lower than among NSUs: 10.3% vs. 15.7% after 30 days, and 16.8% vs. 22.4% after 90 days, corresponding to adjusted 30-day and 90-day mortality rate ratios of 0.69 (95% CI  0.58–0.82) and 0.75 (95% CI  0.65–0.86)There were no statistically significant differences between SUs and NSUs regarding the adjusted RR for bacteraemia (1.07; 95% CI  0.69–1.67) and for pulmonary complications (0.69; 95% CI  0.42–1.14)
 Propensity score technique was used SUs: 1346
NSUs: 1346
Crude and fully adjusted mortality-related ratios in the matched analysis were, respectively, 0.63 (95% CI  0.51–0.78) and 0.64 (95% CI  0.52–0.80) after 30 days, and 0.69 (95% CI  0.58–0.82) and 0.71 (95% CI  0.60–0.84) after 90 days 
[20]Retrospective national cohort study using the VA Department administrative data
Propensity score technique was used
8652 subjects aged >65 years hospitalized with CAP, and having >1 year of prior VA outpatient careSUs: 1567
(SUs only, 798; and SUs combined with ACEiUs, 769)
NSUs: 7085
ACEiUs: 2930 (ACEiUs only, 2161)
Statin use was significantly associated with decreased 30-day mortality (5.0% vs. 10.9%, p <0.001)
After adjusting for the appropriate propensity score and hospital admission, prior use of only a statin (OR 0.58; 95% CI  0.42–0.80) was significantly associated with decreased 30-day mortality
ACE inhibitor use (7.8% vs. 11.0%, p <0.0001) was also significantly associated with decreased 30-day mortality
Non-statin lipid-lowering medication use had no significant association with mortality
[21]Prospective observational cohort study1007 unselected patients admitted to the hospital with radiologically confirmed CAPSUs: 257
NSUs: 750
On multivariate logistic regression, statin use was associated with significantly lower 30-day mortality (adjusted OR 0.46; 95% CI  0.25–0.85; p 0.01) and development of complicated pneumonia (adjusted OR 0.44; 95% CI  0.25–0.79; p 0.006)Statin use had no protective effect on requirement for mechanical ventilation or inotropic support (adjusted OR 0.93; 95% CI  0.49–1.76; p 0.81)
SUs had more severe CAP (median PSI 4) than NSUs (median PSI 3; p <0.0001) but lower admission CRP levels (median 119 mg/L vs. 182 mg/L; p <0.0001)
[22]Retrospective, population-based, nested case–control study134 262 adults aged 30 years or older enlisted in the UK General Practice Research Database
1253 patients with CAP and 4838 controls
SUs: 1407
NSUs: 4684
Current SUs had lower risk for fatal pneumonia (statin use in 6.7% of cases vs. 12.7% in controls—adjusted OR 0.47; 95% CI  0.25–0.88; p 0.02)Current statin use was associated with slightly but not significantly reduced risks of uncomplicated pneumonia (adjusted OR 0.84; 95% CI  0.61–1.17; p 0.30) and hospitalization for uncomplicated pneumonia (adjusted OR 0.74; 95% CI  0.50–1.08; p 0.12)
[23]Prospective, population-based, cohort study
Propensity score was used
3415 patients admitted with CAPSUs: 325
NSUs: 3090
SUs were less likely, but in a not statistically significant way, to suffer the composite outcome of death or admission to an ICU than were NSUs (15.0% vs. 19.0%,; OR 0.80; p 0.15)More complete adjustment for confounding changed the OR from potential benefit to potential harm (OR 1.10; fully adjusted, including propensity scores, 95% CI  0.76–1.60) for risk of death or ICU admission
  1. ACE, angiotensin-converting enzyme; ACEiU, angiotensin-converting enzyme inhibitor user; CAP, community-acquired pneumonia; CRP, C-reactive protein; CVD, cardiovascular disease; ICU, intensive-care unit; NSU, non-statin user; PSI, Pneumonia Severity Index; RR, relative risk; SU, statin user; VA, Veterans Affairs.

[24]Retrospective case–control study142 175 patients with diabetes mellitus types 1 and 2 enlisted in the UK General Practice Research Database
4719 patients with CAP and 15 322 matched controls
SUs: 368
NSUs: 19 673
Current statin use was associated with a reduced risk of pneumonia (crude OR 0.51; 95% CI  0.37–0.68) (adjusted OR 0.49; 95% CI  0.35–0.69)The protective effect was similar for all statins, and was consistent for both summer and winter seasons
Comparable effects were observed in all comorbidity-related subgroups (with the strongest association in patients with a history of CVD), and the protective effect did not differ substantially between younger (OR 0.56; 95% CI  0.37–0.86) and older patients (>70 years; OR 0.35; 95% CI  0.20–0.61), and was similar for episodes of CAP treated in primary care (OR 0.52; 95% CI  0.36–0.77) and those requiring admission to hospital (OR 0.50; 95% CI  0.28–0.89)
[25]Retrospective cohort study
Propensity score was used
787 patients admitted at two academic tertiary-care hospitalsSUs: 110
NSUs: 677
SUs had a lower 30-day mortality (4.5% vs. 9.0%) than NSUs
After adjusting for the propensity score and processes of care, statin use at presentation was associated with decreased 30-day mortality (OR 0.36; 95% CI  0.14–0.92)
 
Table 3.   Clinical evidence regarding the use of statins in bacteraemia
StudyStudy designPatients’ settingsStudy groupsPrimary outcomeSecondary outcomes
  1. ICU, intensive-care unit; NSU, non-statin user; SU, statin user; VA, Veterans Affairs.

[26]Observational cohort study based on prospective registration5353 adult patients hospitalized with bactaeremiaSUs: 176
NSUs: 5177
The 30-day mortality in SUs was similar to that in NSUs (20.0% vs. 21.6%; adjusted mortality rate ratio 0.93; 95% CI 0.66–1.30; p 0.66)Among 30-day survivors, SUs experienced a substantially decreased mortality up to 180 days after the bacteraemia (8.4% vs. 17.5%; adjusted mortality rate ratio 0.44; 95% CI  0.24–0.80; p 0.05)
[27]Retrospective cohort study438 bacteraemic patients admitted to a single acute-care general hospitalSUs: 66
NSUs: 372
Overall hospital mortality was lower among SUs (10.6% vs. 23.1%; OR 0.4; 95% CI  0.17–0.9; p 0.022)Bacteraemia-attributable mortality was also lower among SUs (6.1% vs. 18.3%; OR 0.29; 95% CI 0.1–0.82; p 0.014)
[28]Retrospective cohort study388 patients with bacteraemic infections due to aerobic Gram-negative bacilli and Staphylococcus aureus admitted to a single VA Medical CenterSUs: 35
NSUs: 353
SUs experienced a lower hospital mortality rate (6.0% vs. 28.0%; p 0.002)Bacteraemia-attributable mortality was also lower among SUs (3.0% vs. 20.0%; p 0.01)
The ICU admission rate while the patients had bacteraemia was not significantly different between SUs and NSUs (17.1% vs. 19.3%)
Table 4.   Clinical evidence regarding the use of statins for preventing infections in the postoperative setting
StudyStudy designPatients’ settingsStudy groupsPrimary outcomeSecondary outcomes
  1. CABG, coronary artery bypass grafting; NSU, non-statin user; SU, statin user; VA, Veterans Affairs.

[29]Prospective cohort study
Propensity score was estimated
2497 adult patients who underwent isolated CABG
Serious infection morbidity consisted of sepsis syndrome, septic shock, or mediastinitis
SUs: 1835
NSUs: 662
SUs: 654
NSUs: 654
SUs and NSUs had similar rates of serious infections (1.7% vs. 2.5%; adjusted OR 0.68; 95% CI  0.31–1.48; p 0.33) 
[30]Retrospective cohort study1934 patients who underwent CABG, valve surgery or a combination of bothSUs: 1248
NSUs: 686
Preoperative statin use was associated with a significant reduction in the development of infection (adjusted OR 0.67; 95% CI  0.46–0.99; p 0.04)Statin use was not associated with a significant reduction in any individual infection (p >0.08 for all), such as pneumonia, urinary tract infection, bacteraemia, deep sternal wound infection, leg vein harvest site infection, and tracheotomy site infection
[31]Retrospective cohort study10 782 patients who underwent inguinal herniorrhaphy or ventral hernia repair at ten VA Medical Centers in the mid-south region of the USASUs: 1242
NSUs: 9540
Statin use did not affect the risk of wound infection (0.6% vs. 0.5%)Statin use remained significant as an independent predictor of wound haematoma/postoperative bleeding (OR 1.6; 95% CI  1.03–2.44; p 0.04)

Studies examining the use of statins in patients with sepsis or at risk of sepsis

Nine studies were included in this group [10–18]. The randomized controlled trial by Tseng et al. [12] randomized 80 patients with subarachnoid haemorrhage to receive either pravastatin or placebo for up to 14 days, and showed that, although more patients in the pravastatin group experienced sepsis (pneumonia or catheter-related infections), this difference was not statistically significant. It is of interest that statin use seemed to confer an impressive reduction in sepsis-related mortality (6.25% as compared with 71.43%; log-rank test; p <0.001). Although this study had a randomized design, the small number of patients (40 in each group) and the fact that sepsis incidence and the associated mortality were not the primary outcomes preclude firm conclusions.

The other eight studies were cohort studies (three prospective [14,15,18] and five retrospective [10,11,13,16,17]) and included from 53 [13] up to 69 168 [17] patients (41 145 statin users and 44 750 non-statin users). Propensity-matched subcohorts were used in three of the eight cohort studies [14,15,17], and all of these studies reported positive findings. Specifically, statin use was associated with significantly lower numbers of sepsis-related hospitalizations in two of these studies [14,17] and decreased infection-related mortality in one study [15]. As far as the five studies that had a non-matched design are concerned [10,11,13,16,18], statin use was associated with decreased sepsis/infection rates in two studies [13,18] and with no difference in one study [16]; the other two studies did not provide data on this outcome [10,11]. In addition, statin use was associated with decreased overall 30-day or hospital mortality in one study [10], no difference in two studies [11,13], and increased mortality in one study [16]. In this study, hospital mortality was significantly higher in statin-treated patients, even after adjustment for the expected mortality risk defined by the APACHE II score (observed/expected ratio: 1.53/1.17). The authors of the study hypothesized that ‘underlying clinical conditions were insufficiently considered in mortality predictors’. In conclusion, the studies that focused on sepsis and implemented propensity score analysis and the randomized controlled trial reported findings that were at least as positive as those of the studies without matched design.

Studies examining the use of statins in patients with CAP

In the second group, seven studies were included [19–25]; five of them were cohort studies (two prospective [21,23] and three retrospective [19,20,25]) and two were retrospective case–control studies [22,24]. The individual study sample size ranged from 787 [25] up to 29 900 patients [19] (including 9603 statin users and 60 290 non-statin users).

In the only study with a non-matched design, it was found that statin use was associated with significantly lower 30-day mortality [21]. Propensity-matched subcohorts were used in four studies [19,20,23,25]; statin use at presentation was associated with decreased 30-day mortality in three of them [19,20,25] and no difference in one [23]. In addition, in both case–control studies [22,24], it was found that current statin use was associated with a reduced overall risk of pneumonia or fatal pneumonia. Overall, statin use was associated with decreased mortality in five studies [19–22,25] and no difference in one study [23]. In addition, statin use was shown to reduce the risk of complicated parapneumonic effusion and empyema in one study, although it had no effect on the requirements for inotropic support or mechanical ventilation [21], whereas in another study there was no discernible effect on pulmonary complications [19]. In summary, all studies but one examining the effect of statin use on CAP reported positive results, and the study design seemed to have no effect on the results.

Studies examining the use of statins in patients with bacteraemia

In the third group, three studies were included [26–28]; all of them were cohort studies (one prospective [26] and two retrospective [27,28]). Propensity-matched analysis was not used in any of these studies. The individual study sample size ranged from 388 [28] up to 5353 patients [26] (including 277 statin users and 5902 non-statin users). Both hospital mortality and bacteraemia-attributable mortality were lower among statin users in two studies [27,28]. In one of them, the intensive-care unit (ICU) admission rate due to bacteraemia was not significantly different between statin and non-statin users [28]. In the third study, the 30-day mortality was not lower among statin users [26]; however, among the 30-day survivors, statin use was associated—for reasons that are unclear—with substantially decreased mortality up to 180 days after the bacteraemia episode.

Studies examining the use of statins for infection prevention in the postoperative setting

In the fourth group, three studies were included [29–31]; all of them were cohort studies (one prospective [29] and two retrospective [30,31]). The study sample size ranged from 1934 [30] up to 10 782 patients [31] (including 4325 statin users and 10 888 non-statin users). Two of the studies focused on patients who underwent cardiac operations; in the first study, among 2497 adult patients who underwent isolated CABG, statin use did not seem to reduce the incidence of serious infections (e.g. mediastinitis, sepsis syndrome, or septic shock) [29], whereas in the second study, among 1934 patients who underwent CABG, valve surgery, or a combination of both, preoperative statin use was associated with a significant reduction in the development of infection, even though it did not lead to a statistically significant reduction in any individual type of infection [30]. In the largest study, and the only one in this group that used propensity score analysis, patients who underwent inguinal herniorrhaphy or ventral hernia repair did not seem to benefit from statin use as far as the risk of wound infection is concerned [31].

Overall, data regarding mortality as the main outcome were presented in 15 studies; specifically, in six of nine sepsis studies [10–13,15,16,18], in six of seven CAP studies [19–23,25], in three of three bacteraemia studies [26–28], and in none of the three studies conducted in the postoperative setting. Ten of these 15 studies reported decreased mortality in statin users (three of six sepsis studies [10,12,15], five of six CAP studies [19–22,25], and two of three bacteraemia studies [27,28]), four of 15 studies showed no difference in mortality (two of six sepsis studies [11,13], one of six CAP studies [23], and one of three bacteraemia studies [26]), and in one study that included septic ICU patients, there was increased mortality in those patients who received statins [16]. In addition, nine studies (six of nine sepsis studies [12–14,16–18] and all studies conducted in the postoperative setting [29–31]) examined the risk of developing sepsis/infection as the primary outcome. Five of these nine studies showed a decreased risk of sepsis in patients receiving statins (four of six sepsis studies [13,14,17,18] and one of three studies conducted in the postoperative setting [30]), whereas the remaining four studies found no difference [12,16,29,31].

Finally, regarding studies that used propensity score techniques, six of eight studies reported that statin use was associated with favourable outcomes [14,15,17,19,20,25], whereas in the remaining two studies no difference was shown [23,29]. Both case–control studies reported positive results [22,24], whereas the findings of the small randomized controlled trial cannot be safely taken into account, as they represented secondary outcomes [12]. Among the remaining 11 studies that implemented a non-matched design, seven reported positive findings [10,13,18,21,27,28,30], three found no difference in outcomes [11,26,31], and one reported worse outcomes among statin users [16]. Overall, the difference in the study design (matched vs. non-matched) did not seem to have an effect on the reported findings.

Randomized controlled studies (unpublished and/or ongoing) examining the use of statins in sepsis

We identified seven studies relevant to our review [37–43] in the online registries of RCTs (http://www.clinicaltrials.gov and http://www.controlled-trials.com). Two of them have been completed [37,38], but their findings have not been published yet, and the remaining five trials are still ongoing (four of them are currently recruiting patients [39–43]).

Discussion

In this review, we attempted to identify, and present a synthesis of, the available literature regarding the use of statins for the management of septic patients. Approximately two-thirds of the 22 evaluable studies suggested a statistically significant positive effect of statins in patients with sepsis and infection (as exemplified by decreased mortality rates in ten of 15 studies and decreased risk of infection in five of nine studies). Although our review cannot be considered to be immune to publication bias, it includes all recently published studies and corroborates the findings of previously conducted reviews on the same topic [9].

The major limitation of this review is the inclusion—with one exception [12]—of only observational cohort studies, in which statins were administered for cardiovascular system protection but not for the treatment of an infection. As previously noted, even in the one RCT [12] in which statins were, in a sense, administered for treatment, as they were started in the hospital setting, the primary endpoints were the functional/neurological outcomes of patients with subarachnoid haemorrhage, not the infection rate and its associated mortality. In addition, these studies were markedly heterogeneous regarding: (i) their design (matched (RCT, case–control and propensity-matched cohorts) vs. non-matched cohorts); (ii) the sample size (i.e. 53 vs. 69 168 patients); (iii) the characteristics of the included patient populations (i.e. healthy population-based vs. chronic haemodialysis patients); (iv) the method used for data abstraction (i.e. direct chart review vs. mining computerized databases); (v) the definition used for current statin use (i.e. prescription filled in the last 30 days vs. 365 days before the septic episode, and the validation or not of compliance with statin therapy); (vi) the definition used for each infection (i.e. clinically diagnosed vs. radiologically confirmed pneumonia); (vii) type and severity of the infection under study (i.e. pneumonia vs. wound infection or patients treated in the community vs. ward-admitted patients); (viii) the implemented therapies (i.e. adequacy of antibiotic coverage); and (ix) the reported outcomes (i.e. hospital mortality vs. ICU admission rate). It is of interest that the design of the included studies (matched vs. non-matched) did not seem to explain the heterogeneity of the reported findings. The multidimensional heterogeneity of the reviewed studies did not allow us to perform a formal meta-analysis and draw firm conclusions. This does not mean that beneficial effects of statins on sepsis can be easily written off.

The authors of the individual studies were aware of the limitations of their analyses, and employed different techniques to overcome them. For example, some of the cohort studies used prospective data collection, but their findings did not differ from those of retrospective cohort studies. The most widely discussed limitation of the observational studies exploring the role of statins in sepsis is the confounding by indication for treatment, alternatively called the ‘healthy user effect’ [23,44,45]. This is actually a selection bias; plainly speaking, statin users are not the same as non-statin users. It is reasonable to hypothesize that statin use is more likely among health-conscious patients who have healthy diets, exercise regularly, receive flu shots, do not stop smoking, keep their appointments with doctors, and adhere to prescribed therapies [45]. On the other hand, it should not be forgotten that statin prescription is not merely a lifestyle intervention; statin users usually demonstrate higher cardiovascular morbidity, and this may influence their outcome in a negative way during an infectious complication.

The authors of some studies attempted to control for the healthy user effect by using propensity scoring. In this statistical technique, which was developed 25 years ago [46], multivariable logistic regression analysis is used to estimate the probability of use of an intervention (e.g. insertion of Swan–Ganz catheters) or a drug (e.g. statin) on the basis of preselected patient characteristics. The propensity score is the probability of exposure to this intervention, given the observed parameters, and it may reduce selection bias in trials with non-randomized designs. It is of note that approximately two-thirds of the studies that were included in our review and used propensity-matched subcohorts reported that statin use was associated with positive outcomes.

As propensity scoring cannot fully rule out the possibility of confounding by unmeasured factors or replace randomization, large RCTs have been designed to better define the effects of statins on sepsis incidence and outcomes. The recent paradigms of hormone replacement therapy in postmenopausal women [47] and antioxidant supplementation for cancer prevention [48] that—despite promising data from observational studies—failed to show efficacy in RCTs serve as useful reminders. As previously mentioned, two randomized double-blind placebo-controlled trials have already been completed, and await publication; the first explored whether the administration of simvastatin to ICU-admitted patients with severe sepsis reduced plasma concentrations of interleukin-6 and improved clinical endpoints [37], and the second explored whether the use of atorvastatin in ward patients with sepsis can significantly reduce the rates of conversion from sepsis to severe sepsis and other ‘hard’ clinical outcomes [38]. Five additional studies have still to be completed [39–43], confirming the keen interest of the pharmaceutical industry and the medical community in this issue.

Notwithstanding the merits of the expected RCTs, we think that their findings, either positive or negative, should be viewed with caution for several reasons. First, RCTs are not immune to unequal (chance-dependent) distribution of variables between the intervention and the control groups. Second, not all statins have been created equal, and their anti-inflammatory properties are known to vary, at least in animals [49]. Third, few preclinical studies have tested their administration after the infectious challenge, and the corresponding results are contradictory [49,50]. In that regard, one might consider numerous and costly RCTs to be premature and unjustified. Fourth, statins may not be effective among all patient subpopulations. A 70-year-old man with underlying coronary artery disease admitted to the ICU with pneumonia, acute respiratory distress syndrome and septic shock may not derive the same benefit from statin use as an otherwise healthy 55-year-old woman with pyelonephritis. Fifth, the erratic and unpredictable gastrointestinal drug absorption in critically ill patients, combined with the absence (so far) of intravenous statin formulations, will seriously limit their applicability in the ICU setting [51]. Finally, critically ill patients may be at a heightened risk of statin-associated myopathy and hepatotoxicity resulting from pre-existing comorbidities, intervening organ failures (e.g. renal failure), and polypharmacy (e.g. steroids, macrolides, amiodarone, and muscle relaxants) [51]. These adverse events may easily shift the therapeutic pendulum from a potentially positive effect to a harmful complication.

Conclusion

The last decade has witnessed an increasing interest in the use of statins for the prevention and/or treatment of infections. The majority of the available literature lends support to a potentially positive effect of statin administration. Unfortunately, it is mostly of an observational nature, and therefore may suffer from selection bias and hidden confounding. The harsh reality is that there is no strong clinical evidence to support the routine use of statins in septic patients. However, before throwing the baby out with the bathwater, we should await the results of large RCTs yielding clinically exploitable conclusions.

Transparency Declaration

The authors declare no funding and no conflict of interest.

Ancillary