European Journal of Clinical Pharmacology

, Volume 68, Issue 5, pp 747–755

Statins and associated risk of pneumonia: a systematic review and meta-analysis of observational studies

Authors

  • Chun Shing Kwok
    • School of MedicineUniversity of East Anglia
  • Jessica Ka-Yan Yeong
    • School of MedicineUniversity of East Anglia
  • Richard M. Turner
    • Norfolk and Norwich University HospitalColney Lane
  • Rodrigo Cavallazzi
    • Department of MedicineUniversity of Louisville
  • Sonal Singh
    • Department of MedicineJohns Hopkins University School of Medicine
    • School of MedicineUniversity of East Anglia
Pharmacoepidemiology and Prescription

DOI: 10.1007/s00228-011-1159-4

Cite this article as:
Kwok, C.S., Yeong, J.K., Turner, R.M. et al. Eur J Clin Pharmacol (2012) 68: 747. doi:10.1007/s00228-011-1159-4

Abstract

Purpose

Statins have potential anti-inflammatory effects, but the association between statin use and lower incidence of pneumonia is unclear. We have therefore performed a systematic review on the risk of pneumonia in statin users versus non-users.

Methods

MEDLINE and EMBASE were searched in December 2010 for controlled observational studies that reported on the risk of pneumonia in statin users. We performed a random effects meta-analysis and assessed heterogeneity using the I2 statistic.

Results

A total of 451 citations were screened, and ultimately nine studies (4 case–control, 4 retrospective cohort, 1 prospective cohort) with more than 3 million participants were included in the meta-analysis. Pooled analysis of seven studies that reported unadjusted data failed to show a significantly reduced risk of pneumonia [odds ratio (OR) 0.94, 95% confidence interval (CI) 0.84–1.06, p = 0.33, I2 = 79%] in statin users as compared to non-users. However, a significant reduction in the likelihood of pneumonia associated with statin use (n = 8 studies, OR 0.85, 95% CI 0.75–0.97, p = 0.02, I2 = 81%) was found in the meta-analysis of adjusted data. Both analyses were limited by substantial statistical heterogeneity. Sensitivity analysis failed to fully clarify the source of heterogeneity, but cohort studies seemed to be less heterogenous (n = 5 studies, OR 0.92, 95% CI 0.84–1.01, I2 = 43%).

Conclusion

Our findings indicate that the purported benefit of statins in preventing pneumonia is inconsistent, and of low magnitude, with upper bounds of the confidence interval being close to null. In view of the substantial statistical and clinical heterogeneity in the dataset, there is no convincing evidence to support the therapeutic application of statins for reducing the risk of pneumonia.

Keywords

StatinsPneumoniaMeta-analysis

Introduction

Statins have potential anti-inflammatory, anti-thrombotic and other pleiotropic effects on vascular function [1]. As such, they have been hypothesised to both decrease the risk of pneumonia and favourably affect its course, perhaps through pleiotropic effects, such as anti-inflammatory, anti-oxidant, immunomodulatory, anti-apoptotic, anti-proliferative, anti-thrombotic and endothelium-protecting features that reduce the risk of infection [28]. Statins may also be beneficial in preventing and treating cytokine dysregulation, such as bacterial sepsis [913].

There is considerable debate regarding two potentially useful effects of HMG-CoA reductase inhibitor (statin) therapy, namely, reducing the incidence of pneumonia and improving clinical outcomes in those diagnosed with pneumonia. Two case–control studies [13, 14] found that patients on statin therapy were less likely to develop pneumonia compared with those not on statin therapy. However, more recent studies [1517] have failed to confirm this association, with one study actually reporting a significantly increased incidence of pneumonia in patients using statins [15]. Similarly, the most recent published studies reporting the impact of statins on pneumonia outcomes have been conflicting, with one study demonstrating significant mortality reduction [18] and the other showing no benefit [19]. A recent meta-analysis of statins in patients with severe infection/sepsis showed a significant benefit on mortality after pneumonia, but heterogeneity and weaknesses in study methods were important limitations that precluded any robust conclusions.

However, it is still unclear whether the statin users are less likely to develop pneumonia in the first place. The aim of this study was, therefore, to determine the extent of the benefit (if any) from statin therapy on the risk of incident pneumonia.

Methods

Eligibility criteria

We selected controlled observational (case–control or cohort) studies that reported on the risk of developing pneumonia in statin users as compared to those not receiving statin therapy. The studies included in our analysis had to present a relative measure of association, such as relative risk, odds ratio (OR) and/or hazard ratio or provide sufficient raw data to enable calculation of the effect size. We did not specifically evaluate statin trials because of our previous findings that published clinical trials of pharmacotherapy in diabetes and cardiovascular disease seldom monitor for, or provide details on infective respiratory adverse events that seem to be unrelated to therapy [20].

Search strategy

The MEDLINE and EMBASE database search was carried out by YKL using Ovid SP (December 2010) with no language restrictions, based on the terms:
  • Drug term: ‘(*statin).mp’ OR ‘HMG-CoA-reductase-inhibitor.mp’ OR ‘hydroxylmethylglutaryl-CoA-reductase-inhibitor.mp’

    AND

  • Disease term: ‘pneumonia.mp’ OR ‘lower-respiratory-tract-infection.mp’ OR ‘LRTI.mp’.

In addition, we registered with the PubMed automated electronic notification system to alert us should any new articles appear with the above terms. We also looked at the bibliographies of included studies for additional relevant articles.

Study selection and data extraction

Two reviewers (CSK, YKL) independently and in duplicate assessed all titles and abstracts for studies that met the inclusion criteria and excluded any articles that failed to meet the criteria. Full reports (where available) of potentially relevant trials and studies were retrieved and data independently extracted by two reviewers (CSK or JKY, and RT). Data on study design, drug exposure, study location, participant characteristics and exact nature of outcomes were collected on a spreadsheet and checked by at least one other reviewer (YKL, SS or RC). Where there was uncertainty or discrepancies, the articles were discussed amongst the reviewers and consensus reached. Authors were contacted if there were areas that required clarification.

Assessment for risk of bias

In accordance with the recommendations of the Cochrane Adverse Effects Methods Group, we looked at participant selection, follow-up, ascertainment of exposure and definition and monitoring of adverse outcomes [21].

We planned to evaluate publication bias using the funnel plot if there were more than ten included studies and if there was no significant statistical heterogeneity. [22]

Data analysis

RevMan 5.1.1 (Nordic Cochrane Centre, Kobenhavn, Denmark) was used to conduct a random effects meta-analysis using the inverse variance method for pooled odds ratio and their 95% confidence intervals (95% CI).

We assumed similarity between the risk ratio and odds ratio because adverse pneumonia events were deemed to be uncommon events [23]. Where possible, we extracted both unadjusted and adjusted results so that we could pool both the crude and adjusted risk estimates. The analysis was stratified based on whether the pneumonia was community or hospital acquired.

Statistical heterogeneity

Statistical heterogeneity was assessed using I2 statistic, with I2 values of 30–60% representing a moderate level of heterogeneity [24]. The sensitivity analysis was performed by considering the effect of removing particular studies for analyses that were heterogeneous. The study design, participants and methodology of each study were considered as possible sources of heterogeneity.

Results

The search results yielded nine relevant studies with in excess of 3 million participants. The process of selection is shown in Fig. 1. The included studies were observational studies (1 prospective cohort study, 4 retrospective cohort studies and 4 case–control studies) [1317, 2528]. The main characteristics of the studies and participants are described in Table 1. The outcomes, interventions and quality assessments of the included studies are shown in Table 2. All of the studies were carried out in a community or general practice setting, with the exception of two studies involving hospitalised patients. Only one study was based on a new-user cohort [28].
https://static-content.springer.com/image/art%3A10.1007%2Fs00228-011-1159-4/MediaObjects/228_2011_1159_Fig1_HTML.gif
Fig. 1

Flow diagram of the process of article selection for the meta-analysis

Table 1

Study design and characteristics

Study (first author; year)

Design; country and setting

Number of patients

Mean age

% Male

Inclusion and exclusion criteria for participants

Dublin; 2009 [15]

Community-based matched case-control study; Group Health USA Sept 2000–2002

3,360 participants; 1,125 cases, 2,235 controls

Median 77

51

Age 65–94 years with ≥ 2 years of continuous membership. Two controls for each case selected on age, sex, calendar year and duration of pneumonia-free follow-up. Exclusions: history of recent or serious cancer, immunosuppression, chronic renal failure, nosocomial or aspiration pneumonia.

Fleming; 2010 [16]

Community-based retrospective cohort study; UK; Royal College of General Practitioners Surveillance scheme July 1998–June 2006

329,881 person-years (between 320,000 and 400,000 participants)

53% >65 years

45

Age >45 years with continuous registration with same practice for >4 years. Those with cardiovascular disease, diabetes or hypothyroidism were included. Exclusions: cancer, human immunodeficiency virus (HIV), organ transplant and immunosuppression

Kwong; 2009 [17]

Community-based retrospective cohort study; Canada; admitted to hospital

2,240,638 participants

74

45

Aged >65 years who received an influenza vaccination between 1996 and 2006. Controls matched by age, sex and influenza season

Le Manach; 2011 [25]

Prospective observational study; Hospital surgical register; France; Jan 2001–Dec 2009

1,674 participants; 880 statin therapy, 794 no statin therapy

67

88

All patients who underwent infrarenal aortic reconstructive surgery (aneurysm or occlusive disease of the aorta). Exclusions: post-operative statin withdrawal, emergency surgery

Myles; 2009 [26]

Community-based case–control study; UK; THIN database July 2001–July 2002

25,883 participants

57% >70 years

46

Age >40 years with diagnosis of pneumonia. Six controls matched to each case by practice, sex, age and index date

Rodriguez; 2010 [27]

Cohort study; Spain; hospital inpatients

2,045 participants

69

57

Acute ischemic stroke admitted to stroke unit in a University Hospital

Schlienger; 2007 [13]

Community-based nested case–control study; UK. GPRD database Jan 1995–April 2002

6,091 participants; 1,253 cases, 4,838 controls

73% >60 years

55

Age >30 years enrolled for ≥3 years prior to pneumonia diagnosis. Up to four controls matched based on age, sex and general practice. Exclusions: cancer, HIV or immunosuppression

Smeeth; 2008 [28]

Community-based matched new-user cohort study; UK; THIN database Jan 1995–Dec 2006

729,529 participants, 129,288 new-statin users, 600,241 controls

58% >60 years

50

Aged 40–80 years with >12 months registration with practice, who received first prescription for a statin. Matched controls based on sex, age, general practice and no record of statin prescription prior to index date

van de Garde; 2006 [14]

Community-based retrospective case–control study; UK; patients with diabetes mellitus. GPRD June 1987–Jan 2001

20,041 participants; 4,719 cases, 15,322 controls

73

48

Age >18 years with diabetes mellitus, >12 months registration and newly diagnosed community-acquired pneumonia. Controls were matched for sex, age, general practice and index data

Table 2

Study outcomes, results and risk of bias

Study (first author; year)

Risk factors adjustments

Outcome; ascertainment and validation of pneumonia diagnosis

Definition of use; verification of actual exposure; follow-up period or study duration

Results

Dublin; 2009 [15]

Age, sex, calendar year, comorbidities, oxygen use, smoking, ejection fraction, alcoholism, coronary revascularisation, need for assistance, cognitive function, frailty and medications

Pneumonia; diagnosis through ICD-9 codes and review of chest radiograph reports and hospital records

Current use defined as ≥2 prescriptions for a statin within 180 days before the index date. Based on computer pharmacy database; no specific exposure verification

Fully adjusted risk of pneumonia OR 1.26, 95% CI 1.01-1.56; minimally adjusted OR 1.13 95%, CI 0.95-1.34

Fleming; 2010 [16]

Statins, smoking, sex, age group, practice, study year, medications, pneumococcal vaccinations and comorbidities

Pneumonia and other respiratory infection; diagnosis based on Read codes in electronic dataset; unclear validation

Regular statin user: ≥2 prescriptions of statins prior to 31 December in study year and ≥1 prescription after 31 December. Non-statin user: had no recorded statin prescription prior to 31 December in study year. Electronic prescribing dataset; no specific exposure verification

Adjusted pneumonia OR 0.91, 95% CI 0.73-1.13

Kwong; 2009 [17]

Propensity matching for age, sex, chronic institutionalisation, number of hospitalisations, number of medications, risk factors for influenza complications

Hospitalisation for pneumonia; diagnosis based on database records: unclear validation

Users had ≥1 statin prescription during the 90 days preceding the start of the influenza season (or 90 days preceding the start of the post-influenza season for the post-influenza analysis). Based on electronic database; no specific exposure verification

Pneumonia hospitalisation with propensity score matching OR 0.97, 95% CI 0.94–1.00; crude OR 0.92 95% CI, 0.89–0.95

Le Manach; 2011 [25]

Propensity-matched adjustment for patient and hospital characteristics including age, gender, medical history, surgical characteristics, revised cardiac risk index stratification, preoperative medications

Postoperative pneumonia; diagnosis based on pyrexia, new infiltrate on chest radiograph, leucocytosis, and.microbiological tests

Statin use was based on evaluation by anaesthesiologist 10 days before surgery; no verification of duration of prior use. Statins were reintroduced through nasogastric tube postoperatively the evening after surgery

Postoperative pneumonia propensity adjusted relative risk 0.88 95% CI, 0.62–1.25; crude OR 1.05 95%, CI 0.78–1.42

Myles; 2009 [26]

Comorbidities, smoking, ischaemic heart disease, previous pneumonia, diuretics and nitrates

Pneumonia; diagnosis based on Read codes in primary care records; unclear validation

Current use-most recent statin prescription within 30 days before pneumonia diagnosis. Based on recorded prescriptions on primary care medical records; no specific exposure verification

Adjusted current use OR 0.78, 95% CI 0.65–0.94; crude current use OR 1.04, 95% CI 0.88–1.23

Rodriguez; 2010 [27]

Unadjusted

Pneumonia during hospitalisation; presence of dyspnoea, fever, productive cough, clinical signs and radiological features

History of statin use prior to hospital admission with continued therapy during hospitalisation

Unadjusted statin use 24/306 vs. 177/1739 (OR 0.75, 95% CI 0.48–1.17)

Schlienger; 2007 [13]

Drugs, smoking, BMI, comorbidities, alcoholism, and number of GP visits

Pneumonia based on electronic medical records; unclear validation

Current users-those whose last statin prescription was <30 days before the date of pneumonia diagnosis based on GP database; no specific exposure verification

Adjusted current statin use for pneumonia of any severity: OR 0.71. 95% CI 0.56–0.89

Smeeth; 2008 [28]

Age, sex, propensity score (based on multiple variables), year, comorbidities and medication

Pneumonia based on electronic database records; no specific validation

Classified as users if received first prescription for a statin on or after 1 January 1995, based on computerised records; no specific exposure verification; median follow-up 4.4 years

Fully adjusted risk of pneumonia HR 0.84 95%, CI 0.74–0.95; adjusted for age and sex HR 1.04, 95% CI 0.95–1.14

van de Garde; 2006 [14]

Age, comorbidities, smoking, alcoholism, BMI, drugs, flu and pneumococcal vaccinations and number of general practitioner contacts

Diagnosis of pneumonia on electronic database; no specific validation

Current statin use defined if date of pneumonia occurred during predicted treatment course length of a statin prescription on electronic database; no specific exposure verification

Adjusted OR 0.49 95% CI 0.35–0.69; crude OR 0.51 95% CI 0.37–0.68

BMI, Body mass index; HR, hazard ratio; OR, odds ratio; CI, confidence interval

Most of the included studies were susceptible to some risk of bias owing to the lack of prospective ascertainment of pneumonia using pre-specified monitoring or diagnostic criteria. Hence, pneumonia events were typically recorded based on diagnostic codes in electronic databases. Only one study carried out further validation of the diagnostic coding by evaluation of chest radiographs and checking hospital notes [15], while two studies relied on clinical features and radiographs [25, 27]. Ascertainment of statin exposure was typically through prescription records in the databases. Two studies relied on hospital practitioners who recorded the history of statin use, but neither study reported on the duration or dose of prior statin therapy [25, 27].

Most studies attempted to address confounding through statistical adjustment and/or propensity score matching. Six studies reported both adjusted and unadjusted/minimally adjusted estimates, while two studies reported only adjusted data, [13, 16] and one study provided raw data from which we calculated the crude odds ratio [27].

We judged two studies to have slightly lower risk of bias because there was adequate outcome ascertainment and a wider range of adjustments for potential confounding factors [15, 25].

Pooled unadjusted or minimally adjusted risk of pneumonia with statin therapy

Seven studies were included in the unadjusted risk of pneumonia with statin therapy, of which only two individually showed a significant reduction in likelihood of pneumonia in patients on statin therapy [14, 17]. The pooled results of these studies failed to show an association between statin therapy and reduced risk of pneumonia (OR 0.94, 95% CI 0.84–1.06, p = 0.33, I2 = 79%) (Fig. 2).
https://static-content.springer.com/image/art%3A10.1007%2Fs00228-011-1159-4/MediaObjects/228_2011_1159_Fig2_HTML.gif
Fig. 2

Meta-analysis of unadjusted risk of pneumonia with statin therapy

Pooled adjusted risk of pneumonia with statin therapy

Eight studies evaluated the adjusted risk of pneumonia with statin therapy. Each of the seven studies adjusted for different combinations of demographic variables, such as age, sex, comorbidities, smoking status, alcohol use and medications. Four of the studies showed a significant decrease in pneumonia in patients taking statin therapy. Three studies showed no difference in associated risk of pneumonia with statin therapy, whereas one study showed a significantly increased risk of pneumonia with statin use. The pooled result of all eight studies showed a significant association between statin use and reduction in likelihood of pneumonia (OR 0.85, 95% CI 0.75–0.97, p = 0.02,) (Fig. 3). There was substantial heterogeneity in this analysis (I2 = 81%).
https://static-content.springer.com/image/art%3A10.1007%2Fs00228-011-1159-4/MediaObjects/228_2011_1159_Fig3_HTML.gif
Fig. 3

Meta-analysis of adjusted risk of pneumonia with statin therapy

Sensitivity analysis

Both unadjusted and adjusted analyses were limited by substantial statistical heterogeneity (Table 3). Systematic exclusion of each study and evaluation of the effect on the overall heterogeneity failed to identify a single study which contributed most to the overall heterogeneity. When the studies were stratified by design (case–control or cohort), there was still considerable heterogeneity, but much less so for the cohort studies. The source of the heterogeneity remains unclear but probably reflects the diversity in populations, study design, case and exposure definitions and effects of a multitude of different confounders.
Table 3

Sensitivity analysis of adjusted data from observational studies of statin use and risk of developing pneumonia

Stratification method

Number of studies

OR (95% CI) of pneumonia

Heterogeneity

Random effects

8

0.85 (0.75–0.97)

81%

Fixed effects

8

0.95 (0.92–0.98)

81%

Case-control only, random effects

4

0.78 (0.55–1.09)

88%

Cohort only, random effects

5

0.92 (0.84–1.01)

43%

Excluding diabetes cohort, random effects

7

0.81 (0.70–0.93)

81%

Finally, it must be noted that in two of the studies that appeared to have lower risk of bias (due to a better defined ascertainment of pneumonia and a more thorough adjustment of confounders), statin use was not significantly associated with a reduced risk of pneumonia [15, 25].

Discussion

The overall pooled estimates suggest that there may possibly be a reduction in pneumonia among patients taking statin therapy. However, the estimated relative risk reduction was only around 15%, and the confidence intervals were wide, with the upper bounds of the 95% confidence intervals being close to null. We would also caution against any robust conclusions, given the presence of substantial statistical and clinical heterogeneity in the meta-analysis. In such circumstances, the recommended approach for the meta-analysis is to avoid focusing on the overall estimate and to concentrate instead on assessing the consistency of effects, uncovering treatment modifiers and judging boundary conditions [29].

The inconsistency in findings between the adjusted and unadjusted data is of major concern here. There are a number of potential explanations for the wide variation in study results. Some of the data from these studies are derived from large electronic databases which may have limited sensitivity for important co-morbid conditions [15]. Also, both the exclusion criteria and the factors that were considered for the adjusted analysis were not consistent across the studies. Adequate adjustment for pneumonia-related risk factors is important in reducing healthy-user bias as one previous study of the risk of pneumonia mortality with statin therapy found that the association was no longer significant after a thorough adjustment for confounders (including clinically important risk factors such as impaired mobility and previous smoking history that may not always be available in database records) [30]. Schisterman et al.[31] have warned against the possibility of biasing towards the null if ‘over-adjustment’ is carried out based on intermediate variables on the causal route between drug and outcome. However, careful scrutiny of the statistical adjustments and adjusted odds ratios in the studies that we reviewed failed to show any definitive evidence of over-adjustment (Table 2).

In addition, the results of studies showing a positive association may be limited by a healthy-user bias that could confound the results [14, 16]. Statin therapy may be prescribed preferentially to patients with higher socioeconomic status and a lower risk of pneumonia. Equally, statin therapy may have lowered the risk of cardiovascular events, thus making it less likely for patients to seek medical attention for symptoms such as chest pain and shortness of breath that could have led to an erroneous diagnosis of pneumonia.

On the other hand, statin therapy may also imply greater comorbidity than those not on statin therapy so these patients may be more at risk of pneumonia. Some studies attempted to adjust for selection bias using propensity matching methods [17, 28], but the two studies yielded conflicting findings, with opposing direction of effect (Fig. 3). Furthermore, confounding by unidentified biases relating to healthcare use can never be excluded [16, 28] as the adequacy of adjustment will vary considerably depending on the level of patient detail captured within the databases. Residual confounding and impact of unmeasured confounders are possible given that the databases may not have information on pneumonia risk factors, such as impaired mobility, occupational dust exposure, lung function, among others.

Part of the heterogeneity may stem from one study that found an increased risk of pneumonia in statin users with diabetes mellitus, although exclusion of this study did not eliminate the statistical significance or heterogeneity (Table 3). The increased risk may be due to chance or by confounding by indication, particularly in the hospital admission outcome, as statin use is associated with diabetes and congestive heart failure which may influence the decision for patients to be admitted. Some researchers have postulated mechanisms that interfere with the immune response whereby statins may impair host defence mechanisms that contribute to the spread of infection [15]. This is supported by a study in mice which found that statin-treated mice were more likely to have systemic bacteraemia when inoculated with Klebsiella pneumoniae [32].

Strengths and limitations

There were a few strengths of this review. The analysis included a large overall sample size because the data were collected from epidemiological studies. Furthermore, we did not have any restrictions on study design, which allowed us to consider all available evidence. Also, we considered both unadjusted and adjusted data which show the significance of potential confounders. A variety of sensitivity analyses were performed to assess the role of specific effect modifiers. We used a random effects meta-analysis that takes study heterogeneity into account and distributes weights more evenly between small and large studies, as compared to the fixed effect model.

Our review also has a number of limitations. The evidence here is inconsistent and based on study designs of lower quality with imprecise estimates. There are no data on dose responsiveness of this association, and publication bias cannot be ruled out. While two studies were able to use propensity matching methods to reduce selection bias, we cannot exclude the possibility of healthy-user bias. The quality of the data is dependent on the quality of individual studies which were heterogeneous in terms of methodology. Eight of the nine studies were retrospective, and a number of the large-scale studies relied on data from electronic databases, while other studies reviewed medical records. Ascertainment of exposure may have been based on prescriptions issued, rather than actual medication use as reported by the patient. Validation of database classifications of pneumonia diagnosis is not always certain, given that the level of agreement is around 80% between clinical records for pneumonia and electronic diagnostic codes [33].

Clinical and research implications

The use of statins for several therapeutic applications other than their atheroprotective effect has recently been proposed. While a biological rationale can often be found for the use of statin therapy in a multitude of diseases, caution should be exercised as the evidence for such applications is not always conclusive [34, 35]. Even if there were a genuine effect of statins on pneumonia, it is unclear whether the predicted benefit is sufficient to justify widespread use of medications with recognised adverse effects. For instance, if we make the assumption that statins carry a 15% relative risk reduction, then the number needed to treat (NNT) to prevent one additional episode of pneumonia would be 673 (95% CI 388–2525) based on the annual incidence of pneumonia of 1% in community patients age >65 years [36]. Given the magnitude of this projected NNT, we believe it is unlikely that statin therapy would gain widespread acceptability for prevention of pneumonia, bearing in mind that on average, fewer than one in 12 patients with community-acquired pneumonia will have a fatal outcome at 30 days [37].

The conduct of a clinical trial on the effects of statins on unintended (beneficial or adverse) effects, such as the prevention of pneumonia, is challenging because of the potentially large sample size required. For instance, in a population age >45 years with a background pneumonia incidence of 1.01% [36], we would need 151,000 trial participants to have 90% power to detect a relative risk reduction of 15%. Furthermore, even if statins prove to be effective in reducing the risk of pneumonia, it is unlikely that they will be prescribed solely for this purpose given the low incidence of pneumonia in the general population. Perhaps a more pragmatic approach is to evaluate the protective effect of statins in selected populations that are at a higher risk of acquiring an infection such as those with immune suppression or chronic lung disease.

Conclusions

In conclusion, this meta-analysis has uncovered weak and inconsistent evidence that the risk of pneumonia may be reduced with statin therapy. Furthermore, if there is indeed a benefit associated with statin use, it is not clear whether the absolute benefits are worth the additional resources and potential side effects of therapy. We believe that it may be easier to look into the possibility of designing randomised controlled trials to evaluate statin therapy for reducing mortality in patients admitted to hospital with severe pneumonia.

Acknowledgments

There was no funding source for this study. The authors have no conflicts of interest to declare.

Contributors

CSK, YKL and SS conceptualised the review and developed the protocol. CSK, JKY, RT, YKL, RC and SS abstracted data. CSK and YKL wrote the manuscript. Data was analysed by CSK and YKL. YKL will act as the guarantor for the paper.

Funding

None

Copyright information

© Springer-Verlag 2011