Statins and prevention of infections: systematic review and meta-analysis of data from large randomised placebo controlled trialsBMJ 2011; 343 doi: http://dx.doi.org/10.1136/bmj.d7281 (Published 29 November 2011) Cite this as: BMJ 2011;343:d7281
- Hester L van den Hoek, graduate student13,
- Willem Jan W Bos, internist2,
- Anthonius de Boer, professor of pharmacotherapy1,
- Ewoudt M W van de Garde, clinical pharmacist and epidemiologist13
- 1Division of Pharmacoepidemiology and Clinical Pharmacology, Faculty of Science, Utrecht University, Utrecht, Netherlands
- 2Department of Internal Medicine, St Antonius Hospital, Nieuwegein, Netherlands
- 3Department of Clinical Pharmacy, St Antonius Hospital, Koekoekslaan 1, PO Box 2500, 3430 EM, Nieuwegein, Netherlands
- Correspondence to: E M W van de Garde
- Accepted 3 October 2011
Objective To evaluate whether the potential of statins to lower the risk of infections as published in observational studies is causal.
Design Systematic review and meta-analysis of randomised placebo controlled trials.
Data sources Medline, Embase, and the Cochrane Library.
Study selection Randomised placebo controlled trials of statins (up to 10 March 2011) enrolling a minimum of 100 participants, with follow-up for at least one year.
Data extraction Infection or infection related death.
Results The first study selection yielded 632 trials. After screening of the corresponding abstracts and full text papers, 11 trials totalling 30 947 participants were included. 4655 of the participants (2368 assigned to statins and 2287 assigned to placebo) reported an infection during treatment. Meta-analysis showed no effect of statins on the risk of infections (relative risk 1.00, 95% confidence interval 0.96 to 1.05) or on infection related deaths (0.97, 0.83 to 1.13).
Conclusion These findings do not support the hypothesis that statins reduce the risk of infections. Absence of any evidence for a beneficial effect in large placebo controlled trials reduces the likelihood of a causal effect as reported in observational studies.
Statins (hydroxymethyl glutaryl coenzyme A (HMG-CoA) reductase inhibitors), drugs used to lower cholesterol levels, are prescribed frequently and their cardiovascular benefits are widely accepted in medical practice. In addition to cholesterol lowering effects, statins have anti-inflammatory and immunomodulatory properties, so called pleiotropic effects. Such effects might be beneficial in the management of infections.1 2 3
Whether statins can help to prevent infection is still debated. Since 2006, population based observational studies have reported associations between statins and a reduced risk of infectious outcomes such as pneumonia and sepsis and infection related mortality, with preventive effects up to 76%.4 5 6 7 8 9 10 However, the possibility of biased estimations of effect, which are inherent in observational studies, cannot be excluded. Recently, some investigators suggested that the protective associations reported in many of these studies could reflect bias from “healthy user” effects—that is, that statin users tend to have less severe comorbidity and better functional status than non-users and are more likely to practise other healthy behaviours. Despite extensive efforts to minimise all types of confounding and other biases, the effects of statins on risk of infection are inconclusive.11 12
We hypothesised that if a preventive effect of statins on the risk of infections existed it should be detectable in the data of large placebo controlled trials. Considering the placebo controlled nature of such trials this would mean a true effect. We carried out a systematic review and meta-analysis of placebo controlled trials of statins to evaluate whether the beneficial association between statins and infections published in observational studies is causal.
We carried out a systematic review and meta-analysis of published trials according to the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines.13 A systematic search of Medline in PubMed, Embase, and the Cochrane Library database (the Cochrane central register of controlled trials) for studies reporting an infection as an adverse event or cause of death related to statin use was done from inception to 10 March 2011. To identify further articles we hand searched references and the related citations in PubMed.
We included randomised controlled trials reported in English only. Search terms, MeSH subject headings, and limits were “hydroxymethylglutaryl-CoA reductase inhibitors”, “anticholesteremic agents”, “statin”, “simvastatin”, “rosuvastatin”, “pravastatin”, “atorvastatin”, “fluvastatin”, “cerivastatin”, “pitavastatin”, “lovastatin”, “placebo controlled”, and “randomised controlled trial”.
Inclusion and exclusion criteria
All captured publications were screened according to predefined selection criteria. We included all randomised controlled trials with a statin as active treatment and a placebo as comparator. Eligible studies had a follow-up period of at least 12 months, a minimum of 100 participants, and reported on infections or infection related mortality. We excluded duplicate publications—that is, two or more studies that investigated the same sample.
After the first automated query in the databases, we used a three step process for screening. Firstly, we screened the abstracts. Secondly, we screened the remaining publications on the basis of the corresponding full text publication for reports on adverse events and causes of death. Finally, we screened the articles with information on adverse events and mortality for reports on the occurrence of infections or infection related mortality.
Besides the predefined search strategy, one reviewer (HLvdH) considered trials that had been excluded in the final step for not reporting infections as of potential importance when extensive data on other adverse events were available. The principal investigators of these trials were then contacted for supplementary data.
One reviewer (HLvdH) extracted the data. From each eligible publication data were captured on first author, year of publication, title, journal, study population, primary study outcome, duration of follow-up, number of participants in each group (statin and placebo groups), and type and dose of statin studied. From each included study HLvdH extracted data on the occurrence of infections as adverse events and data on infection related mortality.
Quality and risk of bias assessment
Using the Cochrane risk of bias tool we screened all captured trials for quality of study design, including randomisation, blinding, withdrawal, and dropouts.14 Because we studied unexpected “adverse” events in this systematic review we focused our assessment on quality of reporting adverse events and withdrawal of treatment in both study arms. Withdrawal of treatment was considered to rule out the possibility that differences in denominators may explain differences in calculated infection rates.
Statistical analysis of the dichotomous variables was carried out using relative risk as the summary statistic. We used random effect modelling by DerSimonian-Laird for analysis and reported results in a forest plot with 95% confidence intervals. The relative risks represented the risk of an adverse event happening during study interval in participants taking a statin compared with those taking a placebo. When possible we carried out subgroup analyses and stratified the results for type of statin, statin dosage, patients’ characteristics, and type of infection. A relative risk of less than 1 would favour the statin treated population; we considered the point estimate of the relative risk to be significant at the P<0.05 level if the confidence interval did not include the value 1. Heterogeneity was analysed using the Cochran Q test (χ2) and I2 statistics as a complement to the Q test.15 16 I2 is the proportion of total variation observed between the trials attributable to differences between trials rather than to sampling error (chance), with values less than 30% representing low variation, less than 60% moderate variation, and greater than 60% high variation. Analyses were done with PASW Statistics 18.0 for Windows (SPSS). We carried out additional sensitivity analysis to explore the influence of the quality and risk of bias found in the included trials.
The first database query yielded 607 publications (fig 1⇓). Another 25 studies were identified by hand searching. After removal of duplicates, the remaining 587 articles were screened on the basis of the title and abstract. Only 135 trials fulfilled the inclusion criteria. These trials were further evaluated for reporting on the incidence of adverse events and mortality. Of the 39 trials reporting adverse events and causes of death, 10 publications had information on infections. Additionally, 13 trials were certified as potentially relevant and the authors were contacted for primary and supplementary information (see web extra). Of these 13 trials, only one study group provided additional information. In total, the final selection comprised 11 trials. Analyses were carried out on these 11 trials.
Table 1⇓ summarises the baseline characteristics of the included trials. The 11 trials randomised 30 947 participants: 14 103 (45.6%) received statin therapy and 16 844 (54.4%) received placebo. The average follow-up was 3.3 (range 1-5.1) years. The mean age of the patients was 63 (range 48 to 73) years (see web extra).
Of the 11 randomised controlled trials, one was not double blinded (table 2⇓). Three of the trials stated they were randomised but did not give the specific method of sequence generation, although they did state that investigators and participants were blinded to the process of randomisation. Except for one trial,17 study quality was satisfactory overall and the trials were judged to be at low risk of bias (adequate sequence generation or allocation concealment, double blinding, and clear reporting of withdrawal rates and loss to follow-up). All trials were based on intention to treat analysis of long term treatments. The original publications of the trials included in this meta-analysis reported that 1.1-34.6% of participants assigned to statins had stopped using the drugs by the end of follow-up, similar to the 1.0-44.2% of participants assigned to placebo.
Occurrence of infections
Data on infection related adverse events and infection related mortality were retrieved from the original publications, except for the GISSI–heart failure trial21 where the authors provided the relevant data on request. In total, 4655 patients experienced an infection during treatment, reported as an adverse event or cause of death. Of these patients, 2368 were assigned to statins and 2287 to placebo. Table 3⇓ provides details of the individual trials. Most did not report the site of infection.
Statin treatment and outcome events
In meta-analysis the use of statins was not associated with a decrease in the risk of infection related adverse events compared with placebo (relative risk 1.00, 95% confidence interval 0.96 to 1.05; P=0.93, fig 2⇓). Heterogeneity of the included studies was low (I2=5.5%). Similarly, statins were not associated with a reduction in the risk of infection related mortality (0.97, 0.83 to 1.13; P=0.71, fig 3⇓). The included studies also showed low heterogeneity (I2=18.8%). In sensitivity analysis the exclusion of data from the trial lacking double blinding17 did not significantly alter the results for infection related mortality (0.98, 0.84 to 1.14).
In this systematic review and meta-analysis of data from randomised placebo controlled trials of statins, we found no evidence to support the hypothesis that statins decrease the risk of infection. We pooled the data of 30 947 participants, giving a relative risk of 1.00 for infections and 0.97 for infection related mortality. Absence of any evidence towards a beneficial effect of statins on risk of infection in large placebo controlled statin trials does not support a causal protective association between statins and infections as reported in observational studies.
Comparison with other studies
Our results differ from those of most of the observational studies that examined use of statins and risk of infections. Since 2006, observational studies have reported a possible beneficial effect of statins on risk of infection.43 44 Many such studies showed beneficial effects, but a study by Majumdar et al was the first to introduce the so called healthy user effect as a possible source of bias in these studies. Although these researchers studied the effects of statins on prognosis of infections, their findings that analyses with more thorough adjustment, including for functional status, showed no association might also extend to prevention of infections.11 In 2009 a study reported that statin use was not associated with a decreased risk of pneumonia when adjusted for measures of functional or cognitive state.12 More recently, statins have been linked to an increased risk of infection after stroke and no evidence of an effect on risk of invasive mould infection.45 46
A major methodological difference between these studies and our study may explain the divergent results. Most importantly, and inherent to the study design, observational studies are always at risk of unmeasured confounding. A major strength of our meta-analysis of data from only randomised controlled trials is the absence of such bias. The so called healthy user effect, for example, should be absent when allocation to statin treatment or to placebo is determined by chance. Given that our study showed an effect size of about zero provides evidence that previous findings from observational studies indeed might be biased.
Although it is likely that results from the trials were correct and analyses from observational studies were biased, the reverse scenario is also plausible. Firstly, the reporting of infections was limited and most researchers did not provide additional data on request. Therefore we might have included a non-representative sample of statin trials for infectious outcomes. However, the lack of effect reported in the published trials would be biased only in the unlikely scenario of preferential publication of null findings. Secondly, we cannot rule out the possibility that statins still might be beneficial in some categories of patients. Although our meta-analysis provided no suggestion for a specific subgroup with reduced risk of infection, this might be missed because randomised trials in general include “healthier” people with fewer comorbidities and drugs used concurrently. On the other hand, most of the participants in this meta-analysis had renal disease and diabetes, two known risk factors for infections, and comparable to the type of patients largely present in many of the observational studies showing a large beneficial effect, whereas our study found a null effect. Also, the mean age of the patients in the present meta-analysis did not differ from the available observational studies (63 years v 63 years).
In the present study we focused on the effects of statins on risk of infection. We know that a large amount of observational data exists suggesting acute statin effects to improve the prognosis of infection. In our opinion, prevention and treatment effects are both important, but they are clearly different. It is not inconceivable that statins influence the inflammatory response in the acute phase of infection without affecting the overall risk of acquiring infection. According to Clinicaltrials.gov, several randomised clinical trials of statins in the setting of sepsis and pneumonia are currently under way. The findings from these trials will help to answer what statins can add to our shared effort to improve the prognosis of infection.
Strengths and limitations of the study
Strengths of our study include the effort to minimise confounding and the alternative approach used. Firstly, we carried out a thorough systematic review of all placebo controlled trials of statins for data on the presence of infections. This was followed by a meta-analysis to analyse unbiased statin effects on the risk of acquiring infections. This alternative approach provides a unique chance to explore the heterogeneity among the observational studies on the subject. This approach has also been applied to explore the reporting of statins and their effects on risk of fracture.47
Our study also has some limitations. Mainly, of the 632 publications screened only 11 trials reported on the incidence of infections. Therefore we were unable to carry out subgroup analyses for statin dose, type of statin, patients’ characteristics, and type of infection. Secondly, we have no information on the validity of the infectious outcomes, as these were not predefined study outcomes. We do, however, feel confident that infections are major events that would be noticed. Furthermore, if validity was impaired we would expect this to be non-differential and not to lead to an inaccurate estimate of the relative risk of infection.
Conclusions and policy implications
Our systematic review and meta-analysis of data on infectious outcomes in large placebo controlled statin trials did not provide evidence to support the hypothesis that statins decrease the risk of infections. On the basis of previous observational studies, many investigators have called for randomised trials of statins for the prevention of infections. Our finding of an absence of any protective effect shows that results from the observational studies might be biased and that the arguments for setting up such a trial are weakened. Besides this, given our observed effect size of about zero, the sample size and follow-up time required for adequate power to show a difference would be approximately infinite. Furthermore, such a trial could entail withholding statin treatment for a long period from people with an indication for such treatment. In our opinion, such a study would be unethical. According to Clinicaltrials.gov, numerous randomised clinical trials with long term use of statins are planned or recruiting patients. A better approach could be to push reporting of infectious outcomes in detail whenever statin trials are undertaken. This would have better prospects for identifying subgroups in whom there may be effects worthy of additional testing in a focused randomised trial.
What is already known on this topic
Hydroxymethyl glutaryl coenzyme A reductase inhibitors, or statins, are widely used to prevent and treat cardiovascular disease
Statins have diverse anti-inflammatory and immunomodulatory effects
Observational studies have reported a decreased risk of infections for people taking statins, but biased estimations in these studies cannot be ruled out
What this study adds
In this systematic review and meta-analysis of randomised placebo controlled trials, statins had no effect on the risk of infections
The absence of any evidence towards a beneficial effect of statins on risk of infection in large placebo controlled trials potentially negates the causal effect reported in observational studies
Cite this as: BMJ 2011;343:d7281
Contributors: EMWvdG conceived and designed the study. HLvdH acquired the data, carried out the statistical analysis, and drafted the manuscript. HLvdH had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis and is the guarantor. All authors were involved in interpretation of results and drafting and revising the manuscript. All authors approved the final submitted version.
Funding: No funding sources had a role in the study design, data collection, data analysis, data interpretation, or writing of the report.
Competing interests: All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: no support from any company for the submitted work; WB had relationships with AstraZeneca and Merck/Schering Plough, for participation in the AURORA (A Study to Evaluate the Use of Rosuvastatin in Subjects on Regular Hemodialysis: An Assessment of Survival and Cardiovascular Events) trial and the SHARP (Study of Heart and Renal Protection) trial, respectively, in the previous three years; and no other relationships or activities that could appear to have influenced the submitted work.
Ethical approval: Not required.
Data sharing: No additional data available.
This is an open-access article distributed under the terms of the Creative Commons Attribution Non-commercial License, which permits use, distribution, and reproduction in any medium, provided the original work is properly cited, the use is non commercial and is otherwise in compliance with the license. See: http://creativecommons.org/licenses/by-nc/2.0/ and http://creativecommons.org/licenses/by-nc/2.0/legalcode.