Clinical review: Statins and trauma - a systematic review

Statins, in addition to their lipid-lowering properties, have anti-inflammatory actions. The aim of this review is to evaluate the effect of pre-injury statin use, and statin treatment following injury. MEDLINE, EMBASE, and CENTRAL databases were searched to January 2012 for randomised and observational studies of statins in trauma patients in general, and in patients who have suffered traumatic brain injury, burns, and fractures. Of 985 identified citations, 7 (4 observational studies and 3 randomised controlled trials (RCTs)) met the inclusion criteria. Two studies (both observational) were concerned with trauma patients in general, two with patients who had suffered traumatic brain injury (one observational, one RCT), two with burns patients (one observational, one RCT), and one with fracture healing (RCT). Two of the RCTs relied on surrogate outcome measures. The observational studies were deemed to be at high risk of confounding, and the RCTs at high risk of bias. Three of the observational studies suggested improvements in a number of clinical outcomes in patients taking statins prior to injury (mortality, infection, and septic shock in burns patients; mortality in trauma patients in general; mortality in brain injured patients) whereas one, also of trauma patients in general, showed no difference in mortality or infection, and an increased risk of multi-organ failure. Two of three RCTs on statin treatment in burns patients and brain injured patients showed improvements in E-selectin levels and cognitive function. The third, of patients with radial fractures, showed no acceleration in fracture union. In conclusion, there is some evidence that pre-injury statin use and post-injury statin treatment may have a beneficial effect in patients who have suffered general trauma, traumatic brain injury, and burns. However, these studies are at high risk of confounding and bias, and should be regarded as 'hypothesisgenerating'. A well-designed RCT is required to determine the therapeutic efficacy in improving outcomes in this patient population.

subsequently cholesterol [14]. Th e reduced production of cholesterol in the liver results in the upregulation of low density lipoprotein receptors on hepatocytes with increased capture of circulating cholesterol [15].
Although the benefi cial eff ects of statins have been primarily ascribed to their lipid-lowering properties [14], more recently their anti-infl ammatory actions have been recognised and are now thought to contribute significantly to their disease modifying eff ects [16]. Th at statin's eff ects may include actions beyond cholesterol reduction is suggested by the observation that a range of statin preparations appear to show similar effi cacy with regard to cholesterol reduction in cardiovascular morbidity or mortality, despite diff ering abilities to reduce serum choles terol [17]. In addition, statins have been shown to reduce morbidity in patients who did not have high serum cholesterol or cardiovascular disease, but did have evidence of systemic infl ammation [18]. Th e anti-artherogenic actions of statins include improved endo thelial function [19], with reduced thrombus formation [20] and improved atherotic plaque stability [21], as well as the modulation of infl ammatory responses [16,22,23]. Interestingly, conditions where statins have been found to have a positive eff ect on disease progression or mortality are primarily dependent on leucocyte accu mu lation [24]. Statins may thus promote the timely resolu tion of the infl ammatory response, preventing persistence of infl ammation and resultant pathology.
Th ere is now growing agreement that many of the benefi cial eff ects of statins are dependent on HMG-CoA reductase inhibition. Protein prenylation is required for the normal function of small GTPases such as Rho and this is prevented by HMG-CoA reductase inhibition. In this way statins have wide ranging eff ects on cellular processes such as cell migration, reactive oxygen species generation and secretion of proinfl ammatory cytokines that contribute to the modulation of the infl ammatory response [25], and there is evidence from human and animal studies that these drugs may have benefi cial eff ects in a number of conditions characterised by excessive infl ammation [26,27]. Th ere are also an increasing number of reports of favourable eff ects in patients who have suff ered injury, thought to also be mediated by eff ects on infl ammation and the immune system.
Th e aim of this systematic review is to examine the evidence for benefi cial eff ects of pre-injury statin use and post-injury statin treatment in trauma patients. It will consider trauma in general, selected single-system injuries (fractures and traumatic brain injury), as well as burns patients. It will not consider the broader issues of the eff ect of statin treatment in critically ill patients, although it is acknowledged that a benefi cial eff ect in trauma patients, if present, may be related to improved outcomes from sepsis or acute lung injury.

Research question
Th is review aims to answer the question: 'Does pre-injury statin use or post-injury statin treatment improve mortality, or functional outcomes, or reduce complications in patients who have suff ered trauma in general, traumatic brain injury, burns, or fractures?'

Data sources and search strategy
We searched MEDLINE, EMBASE, and the Cochrane Central Database of Controlled Trials for reports published up to January 2012 to identify studies for inclusion in this review. Th e search was limited to studies involving humans. No language restriction was imposed. Th e search terms included ['hydroxymethylglutaryl-CoA reductase inhibitors' , 'statin' , 'atorvastatin' , 'simvastatin' , 'pravastatin' , 'fl uvastatin' , or 'lovastatin']; and ['fracture healing' or 'fracture'], or ['brain injuries' or 'traumatic' and 'brain' and 'injury'], or ['burn'], or ['wounds and injuries' or 'injury']. Both medical subject heading (MeSH) and free text searches were performed. We also searched the proceedings of trial databases, and the reference lists of identifi ed trials and major reviews.

Study selection
One reviewer (JOJ) examined the titles and abstracts of the electronic search results to identify articles that were obviously irrelevant. Two reviewers (JOJ, and DRT or MJM) then independently examined the full text articles of the remaining studies to determine eligibility. We included both randomised and observational studies, but excluded case series without control groups.

Study appraisal
Given the poor agreement on the use of summary scores and checklists, the quality of the included studies was assessed individually, by methodological domain. For ran dom ised studies, we considered sequence generation, allocation concealment, blinding of participants, personnel, and outcomes assessment, attrition, and selective reporting. For non-randomised studies, we considered the type of comparison, identifi cation of participants, participant allocation, and risk of confounding.

Trauma in general Study characteristics
Our search identifi ed 526 articles, 524 of which were excluded following review of the titles and abstracts. Two articles, both non-randomised studies, were eligible for full-text review, and are summarised in Table 1.
Efron and colleagues [28] conducted a retrospective cohort study based on data from the National Study on the Costs and Outcomes of Trauma (NSCOT) of 18 trauma centres and 51 non-trauma centre hospitals using a complex statistical sampling model. Th e authors used a sample weighting based on the conditional probability of being selected. Th e total weighted number of patients taking statins prior to injury was 529.2 (21.9%), compared with 1,887 (78.1%) who were not prescribed HMG-CoA reductase inhibitors. Pre-injury medication use was docu mented as part of routine care, and abstracted using a mapping program, to ten classes of medication, of which statins formed one. A major limitation of this study is lack of information on type, dose, duration of statin use and compliance. Furthermore, it was not clear whether pre-injury statin use was continued following hospital admission. Efron and colleagues [28] failed to include all adult patients in their analysis, because statin use in patients under 65 was limited only to 1.5% of the population, precluding a meaningful analysis. Th e analysis was therefore limited to subjects aged 65 to 84 years of age. A total of 1,224 such patients were identifi ed for multi variate logistic regression analysis and subgroup analysis for presence/absence of cardiovascular comorbidities [28]. Th e risk of residual confounding is high.
Th e second retrospective cohort study, by Neal and colleagues [29], analysed the data of 295 patients with blunt trauma from the Host Response to Injury Large Scale Collaborative Program. Th ese patients were older (aged 55 to 90 years) with severe blunt trauma, demonstrated by hypotension or hypoperfusion on admission. Casualties with isolated traumatic brain injuries and cervical cord injuries were excluded. Pre-injury medication use was recorded prospectively. Again, this study was also unable to ascertain whether pre-injury statin treatment was continued after admission. A propensity score predicting statin use was created using logistic regression to adjust for baseline diff erences in the cohorts. Cox proportional hazard regression was then used to evaluate the eff ects of pre-injury statin use on mortality, and the development of multiple organ failure and nosocomial infection [29]. Again, the risk of residual confounding is high.

Traumatic brain injury Study characteristics
Our search identifi ed 65 citations relating to traumatic brain injury. Examination of the abstracts categorised 63 as either irrelevant, or as case reports, reviews or editorials. Th ese were excluded from further examination. Two clinical studies, one randomised controlled trial and one retrospective cohort study, underwent full text review [30,31] (Tables 1 and 2).
Th e study by Schneider and colleagues [30] was a retrospective cohort study, based on the collected data from the NSCOT. Th e selection criteria included an abbreviated injury scale of greater than 3 for the head region, and age greater than 65 years, as statin use in patients below this age was rare. Patients presenting with fi xed dilated pupils or who died within 24 hours were excluded, as it was felt unlikely that statins would improve survival in this group. Th e data on pre-morbid medication use were incomplete. A total of 523 individuals met the inclusion criteria, 117 of whom were preinjury statin users. After weighting (to allow for the fact that NSCOT included all patients who died in-hospital, but only a proportion of those who were discharged alive), 965 individuals were represented, 242 of whom were statin users. Outcomes include survival to discharge from hospital, and functional outcome at 3 and 12 months after injury, evaluated using the Extended Glasgow Outcome Scale. Multivariable modifi ed Poisson regression analysis was performed to examine the relation ship between pre-injury statin use and outcomes, accounting for age, new injury severity score, Glasgow Coma Scale, infection, shock, cardiovascular comorbidity (defi ned as a history of myocardial infarction, cerebrovascular or peripheral vascular disease, atrial or ventricular tachyarrhythmias), congestive cardiac failure, gender, beta-blocker use, renal disease, hypertension, mid-line shift on head CT scan, tobacco use, chronic obstructive pulmonary disease, diabetes mellitus and whether the patient was admitted to a trauma centre or not. A signi fi cant interaction between statin use and cardiovascular comorbidity was observed and a term defi ned by this interaction was included in the regression models. Th e interaction between statin use and cardiovascular co morbidity was further examined in analyses that stratifi ed statin users by the presence or absence of underlying cardiovascular disease. Nevertheless, the risk of residual confounding is high, particularly given the unknown validity of the weighting algorithm.
Th e study by Tapia-Perez and colleagues [31], on statin treatment rather than pre-injury statin use, was a double blinded randomised controlled trial of rosuvastatin (20 mg, enterally, once daily) in patients who had suff ered traumatic brain injury. Th e authors categorised the study as a 'pilot' , to evaluate a possible positive eff ect of statin treatment on amnesia and disorientation after head injury. Inclusion criteria were patients aged between 16 and 50 years, with a Glasgow Coma Scale between 9 and 13, and an intracranial lesion demonstrated on CT scan. Th ere were extensive exclusion criteria, including death within 72 hours, or due to causes other than brain injury, head injury within 4 weeks or disability due to neurological or psychiatric disease, multisystem trauma, grade III-IV hypovolaemic shock, surgical management of intracranial lesion, isolated brain stem injury, prior treatment elsewhere, pregnancy, and concomitant medica tion use. Forty-three patients were assessed for eligibility, of whom 22 were randomised. Th e primary outcome was the probability of a positive Galveston Orientation and Amnesia Test score to day 120. Secondary outcomes included cytokine levels at day three, and Disability Rating Scale assessment at three months. Participants and investigators were blinded. Follow-up, to three months, was complete. Analysis was by a Cox regression model. One patient died secondary to complicated abdominal surgery and was subsequently excluded from the analysis. Eight patients received rosuvastatin and 13 received placebo over a period of 10 days post-injury. Th e groups were incompletely matched for neurological criteria as more patients in the rosuvastatin group had absent pupillary refl exes compared to the placebo group, possibly suggesting more severe injury despite Glasgow Coma Scale scores being similar between the groups. Overall, this trial is thus also deemed at high risk of bias.

Outcomes
Schneider and colleagues [30] found that statin users were less likely to die before discharge than non-users (9.1% versus 15.4%, respectively). Multivariate regression demonstrated that, overall, statin users at time of injury had a 76% lower risk of dying before discharge (relative risk (RR) 0.24, 95% CI 0.08 to 0.69). Patient gender and concomitant beta-blocker use were not associated with mortality. Th is protective eff ect was greater in those without pre-injury cardiovascular morbidity; thus, statin use was associated with an 83% decrease in the risk of dying (RR 0.17, 95% CI 0.05 to 0.63). In contrast, there was no demonstrated protective eff ect in statin users with pre-injury cardiovascular morbidity (RR 0.87, 95% CI 0.05 to 1.50). In terms of functional recovery, as determined by extended Glasgow Outcome Scale, statin use was shown to be associated with benefi t at 12 months post-injury (RR 1.13, 95% CI 1.01 to 1.26), although no benefi t was demonstrated at three months post-injury (RR 0.77, 95% CI 0.42 to 1.41), regardless of patients' pre-injury cardiovascular conditions. Tapia-Perez and colleagues [31] demonstrated that rosuvastatin 20 mg per day (n = 8) improved Galveston Orientation and Amnesia Test score (HR 53.76, 95% CI 1.58 to 1,824.64) compared with placebo (n = 13). The effect was strongest for leftsided lesions, which may be related to memory and facial recognition being asso ciated with right-sided cerebral structures. There was no difference in the Disability Rating Scale at three months between the two groups. This study indicates that statin given to victims after severe head injury may reduce the risk of developing amnesia.

Study characteristics
Our search identifi ed three articles, one of which was excluded following abstract review, leaving two full-text articles: one is an observational study, and the other a randomised controlled trial (Tables 1 and 2). We also identifi ed one further, completed, randomised controlled trial, which has, however, not been published. Fogerty and colleagues [32] conducted a retrospective cohort study of 223 consecutive burns patients, aged 55 years and over, admitted to a single regional burn centre, over a period of three years. Continuation of statin use following hospitalization was evaluated, and found to be 77%. Multivariate regression analysis was used to determine odds ratios of in-hospital mortality and septic shock, by pre-injury statin use, adjusting for cardiovascular comorbidities [32]. The study was susceptible to residual confounding.
Akcay and colleagues [33] conducted a prospective randomised controlled trial of atorvastatin treatment (20 mg, enterally, once daily, for 14 days) in 20 patients with burns who had not previously taken statins. Th e primary endpoint was plasma E-selectin, which the authors used as a biomarker of burn injury severity [33]. Patients were followed up to 14 days. Th e study was small, with no information on randomisation or allocation concealment, resulting in diffi culties in assessing the quality of its methodology. Th is study is therefore also at high risk of bias.

Outcomes
Fogerty and colleagues [32] demonstrated decreased odds of in-hospital death in patients taking statins preinjury (OR 0.17, 95% CI 0.05 to 0.57). Th is survival benefi t remained unchanged when stratifi ed by cardiovascular comorbidities. However, the authors failed to demonstrate a benefi cial eff ect of pre-injury statin use on infection (OR 0.90, 95% CI 0.48 to 1.7) or septic shock (OR 0.50, 95% CI 0.20 to 1.30) [32]. Akcay and colleagues [33] demonstrated that, following 14 days' treatment, plasma E-selectin levels were signifi cantly lower in the atorvastatin group compared with the placebo group (10.86 ng/ml versus 21.69 ng/ml, P < 0.05).

Fracture healing Study characteristics
Systematic searching for clinical studies identifi ed 391 publications, of which 371 were excluded as irrelevant following review of the abstracts. Twenty articles underwent full text review. Four were excluded on methodological grounds. Of the 16 studies, 15 were concerned with fracture prevention, examining the eff ect of prior statin treatment on the risk of fractures in general, or certain types of fractures (in particular, hip fractures). All of these studies related to individuals at increased risk, such as older women with osteoporosis, or patients on haemodialysis. Nine retro spective studies demonstrated a benefi t to statin use [34][35][36][37][38][39][40][41][42]. Two secondary analyses, of the LIPID and TIMI-22 trials, however, did not show a reduction in the risk of fracture in patients who had been randomised to statin treatment [43,44]. A dose-fi nding trial also failed to demonstrate a diff erence in bone mineral density in patients taking diff erent doses of atorvastatin [45]. It is diffi cult to know whether the fi ndings of these preventa tive studies can be extrapolated to the acute treatment of patients with injuries, and will therefore not be considered further in this review.
Th ere is a single, randomised, double-blind, parallel, controlled trial reporting the eff ect of treatment with 20 mg simvastatin daily for 12 weeks on the time to distal radial fracture healing using dual-energy X-ray absorptiometry assessment of bone mineral density at 2 and 12 weeks post-injury (Table 2) [46]. Th ere were extensive exclusion criteria. Sixty-two eligible patients were recruited (31 in each arm), although 18 were lost to followup and subsequently excluded, as only a per-protocol analysis was performed. Th is trial has a high risk of bias.

Outcomes
Simastatin 20 mg given post-fracture for 12 weeks did not improve fracture healing as the mean time to fracture union (P = 0.648) and the mean percentage trabecular healing at 12 weeks (P = 0.431) were the same in both groups [46]. Th e low dose of simvastatin used in this study may have contributed to the lack of eff ect, given that other studies of the non-lipid-lowering properties of statins have utilised higher doses [26,27].

Discussion
Th e diagnosis 'trauma' encompasses a wide variety of direct injuries, associated complications, such as infection, and the consequences of treatment, such as lung injury caused by ventilation or blood transfusion. Patients who have suff ered injury therefore constitute a highly heterogeneous group. Nevertheless, there are underlying mechanistic similarities that may be amenable to therapeutic intervention. Trauma is characterised by infl ammation, both local and systemic, and the degree of infl ammation is proportional to the magnitude of the combined insult of injury and treatment. A recent review of the epidemiology of civilian trauma deaths has shown that central nervous system injury continues to be the predominant cause of death after trauma (21.6 to 71.5%), followed by exsanguination (12.5 to 26.6%), sepsis (3.1 to 17%) and multi-organ failure (1.6 to 9%) [47]. Over the past decade, a decrease (15 to 25%) in the proportion of haemorrhage-related deaths has been observed [47], in contrast to deaths due to traumatic brain injury, sepsis, and multi-organ failure, which have not decreased. Th ese conditions are strongly related to excessive and injurious infl ammation. Treatments to control or modulate the infl ammatory response could help to mitigate against compli cations associated with these conditions. HMG-CoA reductase inhibitors are potentially attractive agents because they are generally safe and well tolerated drugs. As a result, treatment with statins is being investigated in a number of infective and infl ammatory conditions, particularly in the critical care setting [48,49]. It is therefore not surprising that the possible eff ect of these drugs in trauma patients has also been considered.
Th is review has considered four separate groups of trauma patients, in whom pre-injury statin use, or postinjury statin treatment, has been evaluated. In total, there are only seven studies, four of which are observational. Th ree of these studies suggest improvements in a variety of clinical outcomes (mortality, infection, septic shock in burns patients; mortality in trauma patients in general; mortality in brain injured patients) in patients taking statins prior to injury, whereas one showed no diff erence in mortality or infection, and an increased risk of multiorgan failure, in general trauma patients. Th e design of these studies makes them susceptible to confounding, and in particular, the 'healthy user eff ect' . Further limitations were the age range and comorbidity of the participants, which was determined by the primary indication for commencing statin treatment. As a result, none of the observational studies included patients below the age of 55, and few below the age of 65.
Th e divergent results reported by the two non-randomised studies on trauma patients in general [28,29] are particularly intriguing. Th e diff erences may be explained by the inclusion criteria, suggesting that statins may exert diff erent eff ects in specifi c groups of patients, possibly related to the type and magnitude of the infl ammatory response present.
Several studies have described a 'rebound' eff ect from 'statin-withdrawal' after prolonged use, resulting in increased mortality after acute coronary syndrome [50], increased myonecrosis after vascular surgery [51], and an increased risk of perioperative cardiac events after discontinuation [52]. Neither of the studies of statin use in trauma patients in general [28,29] were able to ascertain whether pre-injury statin use was continued after admission to hospital. Th is important confounder is therefore unaccounted for.
Two of three trials, of statin treatment in burns patients and brain injured patients, showed improvements in outcome, but relied on surrogate measures. Th e third study, of patients with fractures, demonstrated no accelera tion in fracture union. All three trials had limitations, and are considered to be at high risk of bias. Again, there is insuffi cient evidence to support a specifi c type of statin or dose.
None of the studies included in this review reported adverse eff ects. Statin treatment for hypercholesterolaemia, in the outpatient setting, has been shown to be associated with few complications. However, statins can cause rhabdomyolysis, and it is not known whether trauma patients, or those with hypoperfusion, are more prone to this problem [53]. Statins are also known to have eff ects on the coagulation and fi brinolytic systems, which may underlie some of their protective eff ects in cardiovascular disease [54]. Consideration must therefore be given to the possibility of statins exacerbating acute traumatic coagulopathy. Th e eff ects of statins on haemostasis include diminished tissue factor activity, thrombin generation, and thrombin activity, reduced fi brinogen levels, and alterations in tissue plasminogen activator and plasminogen activitor inhibitor (PAI-1) activity [55,56]. Statins have also been shown to reduce platelet activity [57][58][59]. However, statins have not been reported to be associated with clinically signifi cant coagulopathy in patients with trauma.

Conclusion
Th ere is some evidence from observational studies and small, randomised, prospective trials that statin treatment post-injury, or prior statin treatment, may have a benefi cial eff ect in patients who have suff ered general trauma, traumatic brain injury, or burns. However, these studies are at high risk of confounding and bias, and should be regarded as 'hypothesis-generating' . Animal studies may help to provide further proof of concept, but have intrinsic limitations. Ultimately, only a welldesigned clinical trial will be able to answer the question of whether trauma patients may benefi t from adjunctive HMG-CoA reductase inhibition. Designing such a trial will be challenging, and initial work should perhaps focus on statin-naïve patients who are likely to develop a pronounced infl ammatory response. Th e diffi culties in defi ning such a group are acknowledged. Th e possible role of statins in other conditions -in particular, severe sepsis and acute respiratory distress syndrome -are the subject of several ongoing, well-designed, large-scale, random ised controlled trials. Th e results of these studies are eagerly awaited, as these conditions also aff ect many severely injured patients.