Introduction
Chronic obstructive pulmonary disease (COPD) is a pulmonary disease characterized by chronic airflow limitation, often accompanied by systemic inflammation and multiple organ co-morbidities [
1,
2]. Key medications for treating stable COPD are long- and short-acting bronchodilators. In addition, chronic use of inhaled corticosteroids for patients with advanced COPD and short-course systemic corticosteroids during infective exacerbations of COPD are also commonly prescribed for controlling bronchial inflammation [
1]. In the last decade, statins, which are known to inhibit endogenous cholesterol synthesis in hepatocytes by blocking the synthesis of cholesterol [
3], have repeatedly been reported to have anti-inflammatory actions and to reduce inflammatory markers such as C-reactive protein, interleukin-6, interleukin-8, and tumor necrotizing factor alpha in COPD patients [
2]. A number of recent observational studies have also suggested that statins reduce exacerbation, lung cancer, lung function decline over time, cardiovascular events, and even the mortality of COPD patients [
3‐
19]. A commonly proposed hypothesis is that the anti-inflammatory effect of statins prevents COPD exacerbation, cancer, and lung function decline, which together contribute to better prognosis.
COPD is now the fourth leading cause of death in developed countries [
1]. If existing agents, such as statins, actually prevent death from COPD, millions of patients will benefit, because even the current first choice medications modestly reduce the mortality of COPD [
20,
21]. Three systematic reviews were conducted in 2009 to evaluate the effect of statins on the morbidity and mortality of COPD patients [
6,
7,
22]. However, these systematic reviews did not report on the pooled value for mortality, because only a limited number of original articles existed in 2009, and because these original studies reported outcomes using a variety of measurements such as hazard ratio (HR), odds ratio, and relative risk. Furthermore, no previously published systematic review has sufficiently evaluated the publication bias. Additional studies on this topic have been published in the last five years, and an updated systematic review and meta-analysis has been anticipated. Therefore, the aim of the current systematic review and meta-analysis is to estimate the precise impact of statins on mortality in COPD patients.
Discussion
In the current meta-analysis, the HR of statins for all-cause mortality in nine cohorts presented in eight articles was 0.81 (Figure
2). The meta-analysis showed moderate heterogeneity, probably due to the publication bias. The HR was slightly increased to 0.83 after three possibly unpublished cohorts were imputed (Figures
3 and
4). The HR of 0.83 seems a much more reserved value compared to most of the original studies. However, the HR for all-cause mortality of 0.83 is still compatible with or even better than the treatment effect by bronchodilators for COPD patients [
20,
21] and the treatment effect by statins for high-risk primary prevention cohorts [
29].
During the study search for the current meta-analysis, we found some articles reporting the effect of statins on all-cause mortality, which we could not include in the analysis due to lack of data on HR. In 2006, Mancini conducted a time-matched nested case-control study of two population-based retrospective cohorts and reported that fully adjusted risk ratios of statins for death ranged from 0.49 to 0.53, depending on the cohort definition [
30]. The following year, Ishida researched the correlation between COPD mortality and statin use as expressed by statin sales per capita in an elderly population. Among 47 prefectures in Japan, the correlation coefficient was -0.574 (P < 0.001) [
5]. According to Mortensen’s retrospective cohort study with 11212 patients in 2009, current statin use was associated with decreased 90-day mortality with an adjusted odds ratio of 0.51 [
31]. In 2012, Lahousse reported that long-term (>2 year) statin use was associated with a 39% reduction of all-cause mortality after adjusting for confounding variables. Despite the demonstrable results, the report was not included in the current meta-analysis as it does not report in the form of HR, but of odds ratio [
32]. Although we should adopt a careful attitude toward these results due to the limitations of the study design, it is noteworthy that all of the studies above and the original researches included in the current meta-analysis consistently reported a favorable life prognosis by the use of statins. In addition, favorable effects for systemic inflammation, lung mechanics, quality of life, exacerbation, and admission were also confirmed [
2‐
4].
The role of systematic inflammation in COPD has recently been emphasized [
1,
2]. The currently used medications including inhaled corticosteroids are known to reduce symptoms and airflow obstruction. However these medications have limited effect on the natural history of COPD [
1]. On the other hand, the anti-inflammatory effects of statins on pulmonary and systemic inflammation have been repeatedly reported [
2,
3]. In the current analysis, it is reasonable to assume that hypercholesterolemia was more prevalent among statin users compared to non-statin users and that statin users were at high risk of a cardiovascular event and death. However, the existence of hypercholesterolemia and values of cholesterol and/or triglyceride were not always adjusted in the original studies [
10‐
19]. In short, even though statin users may have hypercholesterolemia, the current meta-analysis indicated improved life prognosis by the use of statins. This effect could not be fully explained by cardiovascular event prevention. The anti-inflammatory effect of statins may prevent death along with cardiovascular event prevention. The anti-inflammatory effect of statins was also reported to be relevant to hospitalization, exacerbation, intubation, and decline of lung function in COPD cases [
2‐
4,
7,
22]. This may explain why statins reduce mortality.
A funnel plot and sensitivity analysis suggested the existence of publication bias (Figures
3 and
4). Publication bias is caused by the tendency of researchers and editors to publish the reporting of positive results. While relatively small weighted studies showing favorable results are likely to be published, those showing inconclusive or harmful results are not likely to be published. To avoid selective reporting of trials, RCTs with clinical trial registration is a reasonable solution [
33]. Results from RCTs such as the ongoing “STATins in COPD Exacerbations” are anticipated [
34]. Studies with high quality score and studies focusing on statins have indicated significantly lower HR (Figure
4), though interpretation is difficult. A simple interpretation is that the HR of 0.66 reported by studies with quality scores of 7 or 8 and by studies focusing on statins is more reliable than the HR of 0.88 reported by the other studies with a quality score of 6. Another interpretation is that there was a publication bias or a selective outcome reporting bias. In other words, authors who observed favorable HR by statins tended to make reports concentrating on statins. As mentioned in the results section above, three articles with near-null HR focused on cardiovascular medications and did not provide a detailed description about statins, which resulted in poor quality scores despite the careful study design [
13,
14,
19]. In our opinion, it is difficult to conclude that results from three studies of lower quality (score = 6) not focusing on statins are less reliable than those from the other studies of higher quality (score = 7 or 8) focusing on statins. The important thing is that pooled HR suggested a statin-favorable result even with the low quality studies that suggested near-null HR.
Besides publication bias, our study has some limitations. First, none of the included studies were RCTs, but observational studies. Although meta-analysis with an RCT is usually preferred, meta-analysis with non-RCT studies is commonly accepted and the number of published meta-analyses with observational studies has increased [
35]. That is because an RCT is not always feasible and timely, and because observational studies often yield effects estimates comparable to RCT [
35,
36]. Furthermore, patients satisfying strict inclusion criteria for RCT do not always reflect real world patients with multiple co-morbidities. We believe that the results from the current meta-analysis are trustworthy, because we conducted the current study following guideline for meta-analysis of observational studies [
36]. Second, HR of 0.52, 0.57, and 0.55 for cardiovascular-related, cancer-related, and respiratory-related mortality were not conclusive due to the limited number of available studies. The small number of the included original studies is also a limitation for evaluating all-cause mortality. Third, some may think that the heterogeneity among studies may detract from the reliability of the current meta-analysis. However, we believe that the observed moderate heterogeneity of I
2 = 52% is acceptable for a meta-analysis. Furthermore, the consistency between the results from HRs by fixed-model (HR = 0.81, 95% CI 0.75-0.86) and random-model (HR = 0.75, 95% CI 0.67-0.85) makes the results reliable despite the heterogeneity.
Conclusion
In conclusion, even after possibly unpublished studies were imputed, the pooled HR of statins to all-cause mortality was 0.83 (95% CI: 0.78-0.89, P < 0.001). The HR of 0.83 is a discreet value compared to many previous observational studies, but is still encouraging. HR for cardiovascular-related, cancer-related, and respiratory-related mortality was not sufficiently evaluated in the current analysis as we could not find a sufficient number of original studies dealing with those forms of mortality. Although this meta-analysis and previous original studies have common limitations in their observational nature, these studies have presented meaningful results. The possibly very propitious treatment effect of statins, which may lead to a paradigm shift in the treatment of COPD, should be re-evaluated by a large-scale RCT.
Competing interests
None of the investigators declare any real or perceived conflicts of interest pertaining to the subject of this manuscript.
Authors’ contributions
All authors contributed to the conception, design, data acquisition, analysis, interpretation, drafting, revising, and final approval of the manuscript. NH who is a statistician served as a principal investigator (guarantor). NM, and RK provided interpretation of data and drafting. MI worked for study search. YI, AU, and TK provided study management.