Statins Reduce Mortality after Non-severe but Not after Severe Pneumonia: a Systematic Review and Meta-analysis

PURPOSE: The objective of this study was to perform a systematic review and meta-analysis of the effects of statins on mortality for patients with non-severe pneumonia or severe pneumonia. METHODS: PubMed, EMBASE, Cochrane Database of Systematic Reviews, Cochrane central register of controlled trials and Clinicaltrials.gov were searched for the association between statins and non-severe/severe pneumonia. Eligible articles were analyzed in Stata 12.0. RESULTS: The database search yielded a total of 566 potential publications, 24 studies involving 312,309 patients met the eligibility criteria. Pooled unadjusted data showed that statin use was associated with lower mortality after non-severe pneumonia (odds ratio [OR] 0.70, 95% confidence interval [CI], 0.66-0.73), but not severe pneumonia (OR 1.05; 95% CI, 0.86-1.28). However, this protective effect of statins was weakened using adjusted estimates (OR 0.78, 95% CI, 0.75-0.82). Besides, protective effect of statins was attenuated by confounders in a subgroup analysis, especially when accounting for pneumonia severity indicators (OR 0.88; 95% CI, 0.80-0.96). CONCLUSIONS: Statin use was associated with reduced mortality after non-severe pneumonia but not severe pneumonia and this protective effect was weakened in subgroups.


INTRODUCTION
Pneumonia is one of the highest mortality diseases [1]   .In United States, nearly 4 million adults had community-acquired pneumonia (CAP) per year [2] , which caused almost 50,000 deaths and 1.1 million hospital admissions [3,4] .The incidence of pneumonia was almost 3.3 per 1000 patients for hospitalized patients [5] and the annualized total medical costs reached 14,038 dollars per patient [6] .Although great progress had been made in antimicrobial treatment, mortality of pneumonia was still high, especially for severe pneumonia, mortality rate of which was as high as 50% [7] .
Statins, as one of the inhibitors of 3-hydroxy-3-methylglutaryl coenzyme A reductase (HMG-CoA), are able to lower the level of blood cholesterol and used in patients with cardiovascular diseases or to prevent cardiovascular events [8][9][10] .Stains have potential immunomodulatory and anti-inflammatory effects in CAP [11,12] .An earlier retrospective cohort study showed that, in bacteremia patients, in-hospital mortality was significantly reduced after using statins [13] .Since 2005, researchers have focused more attention on statins in the treatment of infections [14][15][16] .Most of these studies argue that statins are advantageous to outcome and prognosis of patients with infectious diseases.However, Fernandez et al demonstrated that hospital mortality was significantly higher after statin therapy [17] .Majumdar et al reported that statins were not associated with reduced mortality in a prospective cohort study of 3415 patients with pneumonia [18] .Whether statin use was associated with reduced mortality for patients with pneumonia is still in debate.
In this study, we performed systematic review and meta-analysis to address the roles of statins in non-severe and severe pneumonia.Further, subgroup analysis was conducted taking pneumonia severity, propensity score, comorbidity and smoking status as important confounders._________________________________________

Information Sources and Search Strategy
This meta-analysis was conducted according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) and Meta-analysis of Observational Studies in Epidemiology (MOOSE) recommendations [19,20] .PubMed, EMBASE, Cochrane Database of Systematic Reviews, Cochrane central register of controlled trials and Clinicaltrials.govwere searched using the following key terms: (hydroxymethylglutaryl-CoA reductase inhibitors or HMG-CoA reductase inhibitors or simvastatin or lovastatin or pravastatin OR fluvastatin or atorvastatin or rosuvastatin or pitavastatin or statins) and (pneumonia or low-respiratory-tract-infection or lung injury [ALI]or pulmonary injury or acute respiratory distress syndrome [ARDS]).The final date of the literature search was September 31, 2014.There was no publication date, language or status restrictions for searching."Mortality" was not involved in search strategy for reducing the omission.All studies were downloaded into EndNote 6.0 for further screening.Notably, we successfully contacted Bauer to obtain specific data for this meta-analysis.

Study Selection
Studies were included if they met the following criteria: participants should be above 18 years old; they focused on the association between statin use and pneumonia; they reported mortality after an episode of pneumonia; and involved overall or adjusted mortality.We excluded studies that reported ventilator-associated pneumonia (cannot distinguish it from severe pneumonia) and did not report the mortality rate after pneumonia.

Data Extraction
Data were extracted from all included studies independently by Mingwang Jia and Li Li.Relevant information of each study include study design, sample demographics (i.e., age, gender), type of statins, type of pneumonia, type of statins, outcome measures, association scale (odds ratio [OR] and adjusted OR), and adjusted confounders.When disagreement appeared between the 2 reviewers, a senior reviewer (Wenjie Huang) was consulted for final decision.

Quality Assessment
We used Newcastle-Ottawa Quality Assessment Scale (NOS) [21] to assess the quality of observational studies.NOS include three-part: quality of subject selection, comparability between two groups, and reliability of exposure or clinical outcomes.The full score of NOS is 9 and studies were considered of high-quality when scores ≥ 6.We used Jadad Score [22] to assess the quality of clinical trials.Also, Jadad Score include three-part: randomization, blinding, and withdrawals or dropouts.The full Jadad Score is 5 and studies were considered of high-quality when scores ≥3.

Definition of non-severe/Severe Pneumonia And Exposure/Treatment Groups
The Infectious Disease Society of America and American Thoracic Society in 2007 issued consensus guidelines on CAP and SCAP (IDSA/ATS 2007) [23] .Severe pneumonia group was defined as patients need mechanical ventilation or were in the ICU (requiring vasopressors) in our research.Treatment group was defined as patients taking statins (including prior and current users).

STATISTICAL ANALYSIS
We performed all statistical analysis using Stata 12.0 software (StataCorp., College Station, TX).We extracted both unadjusted and adjusted OR for pooling both crude and adjusted risk estimates.OR < 1 indicated that risk of outcome events in the intervention group is lower than that of outcome events in control group, which means that statins reduce mortality of pneumonia.We used I 2 statistic to test heterogeneity of the included studies.An I 2 value of 25%, 50%, and 75% represents low, moderate, and high heterogeneity, respectively [24] .Publication bias was assessed by Begg's test and Egger's test with P < 0.05 as indicative of publication bias.
We performed sensitivity analysis to test the robustness of our findings.The study design and participants of each study were considered as possible sources of heterogeneity.Meanwhile, we conducted an influence analysis for knowing the specific ones which cause the heterogeneity.
For a better understanding of the influenced confounders for mortality, we performed a subgroup analysis.We took pneumonia severity, propensity score, comorbidity and smoking status as important confounders and analyzed the relationship between confounders and mortality.

Subgroup Analysis
We performed subgroup analysis to determine whether clinical factors influenced the mortality of pneumonia after statins using (Table 3).We took pneumonia severity, propensity score, comorbidity and smoking status as important confounders.In the non-severe pneumonia group, no statistically significant differences were observed.But we found that the protective effect weakened when studies were analyzed according to the inclusion of important confounders in their models, especially the adjusted pooled OR for pneumonia severity (0.88, 95% CI 0.80-0.96)and for propensity score (0.86, 95% CI 0.80-0.94),and the more confounders were considered, the weaker the protective effect was.In the severe pneumonia group, though OR was lower in the adjusted data, nonetheless, there was no association between statins users and non-statin users.

Sensitivity analysis
There was a substantial degree of heterogeneity in both unadjusted and adjusted analysis, especially in the non-severe pneumonia group.We performed several sensitivity analyses (Table 4).The I 2 statistic was as low as 0% and 51.5% for non-severe pneumonia (unadjusted) and non-severe pneumonia (adjusted) in prospective cohorts; 28.6% and 53.2% in case-control cohorts, respectively; 0.0% for severe pneumonia in a RCT.The I 2 statistic was as low as 0.0% for severe pneumonia (unadjusted) in North America, 63.7% and 0% for non-severe pneumonia (adjusted) and severe pneumonia (unadjusted) in Europe, respectively.Meanwhile, we performed an influence analysis (Figure 5), which found that, in the unadjusted group, when excluded Kwong's article (only provided the OR value but does not provide a specific number of patients), the I 2 statistic changed into moderate heterogeneity (I 2 = 49.4%,P = 0.04) but didn't change result (OR 0.64; 95% CI, 0.61-0.68).In the adjusted group, we excluded Kwong and Rothberg's articles.The I 2 statistic changed to moderate heterogeneity (I 2 = 41.1%,P = 0.06) but didn't change result (OR 0.69; 95% CI, 0.65-0.74).Articles excluded on the basis of title and abstract N=474

Duplicates articles N=51
Figure 2. Forest for the pooled unadjusted association between statin use and mortality.Forest plot showed that statin users were less likely to die after an episode of non-severe pneumonia compared with nonstatin users in unadjusted estimates, but in the severe pneumonia group, there was no obvious difference in mortality.

DISCUSSION
This meta-analysis of 24 studies showed that statin users were less likely to die after non-severe pneumonia compared with non-statin users, but in the severe pneumonia group, there was no obvious difference in mortality.In the non-severe pneumonia group, statins was associated with reduced mortality using both unadjusted and adjusted estimates.However, this protective effect was weakened in subgroups.
In non-severe pneumonia group, statin reduced the mortality using adjusted and unadjusted estimates but there was no association in severe pneumonia group using either adjusted estimates or unadjusted estimates.Kwok's meta-analysis did not find an association between pneumonia and statins using unadjusted data but found an association using adjusted data [48] .Chopra's meta-analysis found an association between pneumonia and statins using both unadjusted data and adjusted data [49] .These results suggested the severity should be considered when using statin for pneumonia treatment.In a subgroup analysis, we obtained similar results to Chopra [49] .
Figure 3. Forest for the pooled adjusted association between statin use and mortality.Forest plot showed that protective effect of statins was weakened in non-severe pneumonia when accounting for adjusted estimates, and there was no obvious difference in mortality in adjusted severe pneumonia group.
The protective effect we found was weakened when accounting for patient differences through the use of propensity scores, pneumonia severity indicators, smoking status and comorbidity in non-severe pneumonia group.
There were several limitations for this Systematic Review and Meta-analysis.First, the definition of severe pneumonia has flaws.We define patients who need mechanical ventilation or in the ICU (requiring vasopressors) as the severe pneumonia group in our research, and we may have missed severe pneumonia patients who did not use mechanical ventilation or vasopressors or did not mention these parameters in the papers.Additionally, not all ALI/ARDS were caused by pneumonia.However, considering that pneumonia is the main reason of ALI/ARDS and that there must be lung inflammation after developing ALI/ARDS, we define ALI/ARDS as severe pneumonia in our paper.Second, few papers discussed the adjusted OR in severe pneumonia, which may lead to an obvious bias.Thus, we performed a subgroup analysis and analyzed an unadjusted group for better understanding of severe pneumonia.Third, although we conducted adjusted measures and performed subgroup analysis, residual confounding and healthy-user bias remain threats to our conclusions.However, our analysis has important strengths.We divided pneumonia into non-severe pneumonia and severe pneumonia and performed a subgroup analysis to determine whether methodological or clinical factors influenced the meta-analytical estimates of statins on pneumonia mortality.
Statin use is associated with a reduction in mortality in non-severe pneumonia, but because this association in attenuated in studies with a better adjustment for confounders, it is likely that this association is at least partly explained by a healthy user effect.The protective effect of statin was weakened in a subgroup analysis by confounders in non-severe pneumonia, especially when accounting for patient differences through the use of pneumonia severity indicators.In future work, a double-blind, randomized, large sample experiment is necessary considering pneumonia severity as an important element.

Table 2 continued…
Frost; 2007 [41] Duration of enrollment before initiation of statin therapy, the CCI, the number of different medications taken, receiving influenza vaccinations after initiation of statin therapy.

Figure 5 .
Figure 5. Sensitivity analysis for unadjusted non-severe pneumonia (A) and adjusted non-severe pneumonia (B).

Table 1 .
General Characteristics of Included Studies

Table 2 .
Outcomes and Quality of Included Studies

Table 3 .
Subgroup Analysis I= confidence interval; OR= odds ratio; N/A= not available.*Contain one study, can't get I 2 .