- Daniel Caldeira, cardiologist resident, assistant of clinical pharmacology1,
- Joana Alarcão, scientific consultant, assistant of clinical pharmacology2,
- António Vaz-Carneiro, clinical professor of medicine, director of the Center for Evidence-Based Medicine23,
- João Costa, professor of clinical pharmacology, coordinator of the Portuguese Cochrane Centre123
- 1Laboratory of Clinical Pharmacology and Therapeutics, Faculty of Medicine, University of Lisbon
- 2Center for Evidence-Based Medicine, Faculty of Medicine, University of Lisbon, Avenue Professor Egas Moniz, 1649-028, Lisbon, Portugal
- 3Cochrane Coordinating Center Portugal, Faculty of Medicine, University of Lisbon
- Correspondence to: J Costa
- Accepted 10 May 2012
Objective To systematically review longitudinal studies evaluating use of angiotensin converting enzyme (ACE) inhibitors or angiotensin receptor blockers (ARBs) and risk of pneumonia.
Design Systematic review and meta-analysis.
Data sources Medline through PubMed, Web of Science with conference proceedings (inception to June 2011), and US Food and Drug Administration website (June 2011). Systematic reviews and references of retrieved articles were also searched.
Study selection Two reviewers independently selected randomised controlled trials and cohort and case-control studies evaluating the use of ACE inhibitors or ARBs and risk of pneumonia and retrieved characteristics of the studies and data estimates.
Data synthesis The primary outcome was incidence of pneumonia and the secondary outcome was pneumonia related mortality. Subgroup analyses were carried according to baseline morbidities (stroke, heart failure, and chronic kidney disease) and patients’ characteristics (Asian and non-Asian). Pooled estimates of odds ratios and 95% confidence intervals were derived by random effects meta-analysis. Adjusted frequentist indirect comparisons between ACE inhibitors and ARBs were estimated and combined with direct evidence whenever available. Heterogeneity was assessed using the I2 test.
Results 37 eligible studies were included. ACE inhibitors were associated with a significantly reduced risk of pneumonia compared with control treatment (19 studies: odds ratio 0.66, 95% confidence interval 0.55 to 0.80; I2=79%) and ARBs (combined direct and indirect odds ratio estimate 0.69, 0.56 to 0.85). In patients with stroke, the risk of pneumonia was also lower in those treated with ACE inhibitors compared with control treatment (odds ratio 0.46, 0.34 to 0.62) and ARBs (0.42, 0.22 to 0.80). ACE inhibitors were associated with a significantly reduced risk of pneumonia among Asian patients (0.43, 0.34 to 0.54) compared with non-Asian patients (0.82, 0.67 to 1.00; P<0.001). Compared with control treatments, both ACE inhibitors (seven studies: odds ratio 0.73, 0.58 to 0.92; I2=51%) and ARBs (one randomised controlled trial: 0.63, 0.40 to 1.00) were associated with a decrease in pneumonia related mortality, without differences between interventions.
Conclusions The best evidence available points towards a putative protective role of ACE inhibitors but not ARBs in risk of pneumonia. Patient populations that may benefit most are those with previous stroke and Asian patients. ACE inhibitors were also associated with a decrease in pneumonia related mortality, but the data lacked strength.
Pneumonia represents an important clinical condition because of its relatively high incidence (0.5% to 1.1% annually in the United Kingdom) and associated morbidity and mortality.1 2 Susceptibility is higher among elderly people (≥65 years), those with alcohol dependency, smokers, and patients with heart failure, previous stroke, diabetes, chronic kidney disease, and chronic lung disease.3 4 5 6 Pneumonia is a common reason for hospital admission and a risk factor for prolonged hospital stay, carrying a considerable financial burden on healthcare resources.7 8
Usage of some drugs has been shown to modulate the risk of pneumonia. Acid suppressants can increase patients’ susceptibility to pneumonia, whereas statins may have a protective role.9 10 Angiotensin converting enzyme (ACE) inhibitors and angiotensin receptor blockers (ARBs) are often used in patients with cardiovascular disease. ACE inhibitors are known to have adverse effects on the respiratory system, in particular an increased incidence of cough. Basic investigation has shown that bradykinin and substance P sensitise the sensory nerves of the airways and enhance the cough reflex,11 12 13 which may have a protective role on the tracheobronchial tree.14 15 These mechanisms also improve swallowing by avoiding the exposure of the respiratory tree to oropharynx secretions.11 14 16 Taken together, the pleiotropic effects of ACE inhibitors were suggested to reduce the incidence of pneumonia, but available clinical evidence lacks strength17 18 19 and published results have been contradictory.20 21 22
We systematically reviewed and meta-analysed all studies (experimental and observational) evaluating the use of ACE inhibitors and incidence of pneumonia. Because the clinical characteristics and risk factors of populations using ARBs are similar to those of patients using ACE inhibitors, and therefore studies evaluating these interventions share identical potential clinical confounders, we also estimated the incidence of pneumonia in studies evaluating ARBs. Moreover, patients treated with ARBs are less likely to experience respiratory adverse events,23 24 and therefore ARBs may have a protective role.
The systematic review was carried out in accordance with the meta-analysis of observational studies in epidemiology and preferred reporting items for systematic reviews and meta-analyses statements.25 26
Our primary outcome was the incidence of pneumonia. We considered cases of pneumonia, lower respiratory tract infections, and admissions to hospital due to lower respiratory tract infections. Data were extracted irrespective of whether they had been reported as predefined outcomes or as adverse effects. If studies reported data for death from pneumonia only, to avoid duplication we did not consider these cases for the primary outcome. The secondary outcome was pneumonia related mortality, defined as death directly related to this condition or in-hospital death or mortality within 30 days after onset of pneumonia.27 For both outcomes, we did not consider undefined data or data on upper respiratory tract infections.
We considered randomised controlled parallel trials, cohort studies, and case-control studies with ACE inhibitors or ARBs as interventions and with predefined outcomes. Treatment arms could compare ACE inhibitors and ARBs with each other or with placebo or any other active drug. Cohort studies could be based on populations in the community or those in institutions or hospital and had to follow patients to determine pneumonia outcomes.
In case-control studies, cases had to be defined as patients with new onset pneumonia identified through clinical examination, radiological methods, or database codes. Controls had to be matched to cases, but without new onset pneumonia. For pneumonia related mortality, we allowed case-control studies with both cases and controls having pneumonia.
We allowed all participants, irrespective of baseline diseases and risk factors.
Information sources and search method
We identified potentially eligible studies through an electronic search of bibliographic databases from inception to June 2011 (Medline through PubMed and Web of Science with conference proceedings). See the supplementary file for details of the search strategy. No language restrictions were applied. We screened and cross checked identified systematic reviews and meta-analyses evaluating ACE inhibitors or ARBs, as well as reference lists of papers for potential additional studies. We also searched the Food and Drug Administration website (10 June 2011) for regulatory documents with unpublished data from clinical trials.
Study selection and data collection process
The titles and abstracts of obtained records were screened. Doubts and disagreements were resolved by consensus. We assessed the selected studies in full text to determine appropriateness for inclusion. Two authors independently extracted data on study design, location, period of study, patients’ characteristics, drug use and how it was assessed, primary outcomes, data of required outcomes, and adjustments of estimates.
When studies presented different estimates on primary outcomes according to the severity of pneumonia, we extracted for analysis only those reporting the most severe cases. For the primary outcome we considered drug withdrawals due to pneumonia only when no other estimates were available. When more than one risk estimate was available from several sources, we used only the most precise or adjusted measures of association from each report. Otherwise we used the crude odds ratio or derived it from the raw data.
Two authors (DC and JA) independently analysed the quality of reporting by using a qualitative classification according to risk of bias (high, unclear, or low). For observational studies we used a six item classification based on the meta-analysis of observational studies in epidemiology,25 the quality assessment tool for systematic reviews of observational studies,28 and strengthening the reporting of observational studies in epidemiology.29 This system was adapted from a previously published systematic review30 31 and took into consideration the participants (if any justification was given for the cohort and the study reported appropriate inclusion and exclusion criteria), intervention (if drug use was adequately assessed and not based on self report), outcome (if pneumonia was assessed by clinical examination, radiological methods, or database codes and not based on self report), and outcome adjustments (for both age and at least one of the following: smoker or pulmonary disease, cardiovascular diseases or drug use or chronic kidney disease; other adjustments). For randomised controlled trials we adapted the Cochrane Collaboration’s tool for assessing risk of bias to evaluate the quality of reporting: randomisation method, allocation concealment, blinding of participants and staff, blinding of outcome assessment, selective reporting (if pneumonia was a prespecified outcome), and description of withdrawals.32 From these tools we derived risk of bias graphs.
We used RevMan 5.1.4 software for statistical analysis (Nordic Cochrane Centre, Cochrane Collaboration, 2011) and to derive forest plots showing the results of individual studies and pooled analysis.
We carried out three analyses. Firstly, we compared ACE inhibitors and ARBs with each control group using random effects meta-analysis weighted by the inverse variance method to estimate pooled odds ratios and 95% confidence intervals. Heterogeneity was assessed with the I2 test, which measures the percentage of total variation between studies due to heterogeneity.33 We used the random effects model independently of the existence (I2≥50%) of substantial heterogeneity between the results of trials, as we pooled the results of studies with different designs and patients’ characteristics. We chose the odds ratio as the measurement estimate for effect because relative estimates are more similar than absolute effects across studies with different designs, populations, and lengths of follow-up.34 Raw data were first converted to odds ratios through classic methods, or through Peto’s method if one arm had a zero count cell. When raw data or odds ratios were not available we took the hazard ratio or risk ratio for analysis. To explore differences in estimates for outcomes we presented the results stratified according to study design. For the purpose of the analysis we treated nested case-control studies as cohort studies. We carried out subgroup analyses for patients with previous stroke, heart failure, and chronic kidney disease because such patients are known to be particularly susceptible to pneumonia and all indications are clinically approved for treatment with ACE inhibitors and ARBs.5 6 35 36 In view of the suggestion that ACE inhibitors may be more efficient in reducing the risk of pneumonia in Asian patients we also calculated estimates for Asian and non-Asian populations.37 We evaluated differences between subgroups with the method described by Deeks et al, based on the inverse variance method.33
In the second analysis we carried out adjusted indirect comparisons between the pooled estimate of ACE inhibitors (versus control) and ARBs (versus control) using the Bucher frequentist method, which compares different treatments adjusted to the results of their direct comparison with a common control.38 This method partially overcomes the problem of different prognostic characteristics between participants among studies, and it is believed to be valid assuming that the relative effect of interventions is consistent across different studies, as verified in our case.39 By default we used the random effects model because adjusted indirect comparisons that used the fixed effects model tend to underestimate the standard errors of pooled estimates.39
Thirdly, we combined evidence generated by indirect comparisons with evidence from head to head studies comparing ACE inhibitors with ARBs using the random effects model for quantitative pooling,40 41 and we determined the discrepancy and heterogeneity between direct and indirect estimates.42
We also calculated the number needed to treat (NNT) and 95% confidence intervals, taking into account the baseline risk (weighted proportion of event rate in control group) because of the differences in the predicted absolute benefit of treatment according to variation in baseline risk between groups.34 43 In our case, the weighted risk of pneumonia in the control groups was 4.6% (95% confidence interval 3.1% to 6.7%). Publication bias was assessed through visual inspection of the asymmetry in funnel plots.
The search of the electronic databases yielded 807 published studies. After applying the inclusion and exclusion criteria 29 studies were included for analysis (fig 1⇓). The results from eight additional studies were identified in the FDA regulatory documents. Overall, data were obtained from 37 studies.20 21 22 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79
Description of studies
The 37 studies included 18 randomised controlled trials,20 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 11 cohort studies,62 63 64 65 66 67 68 69 70 71 72 73 two nested case-control studies,21 74 and six case-control studies.22 75 76 77 78 79
Among randomised controlled trials, eight were done worldwide,47 48 51 52 55 57 58 61 six in Europe,44 45 46 53 56 60 three in Asia,49 50 59, and one in Europe and the United States.54 Most of the randomised controlled trials were multicentre (n=16). Seven randomised controlled trials compared ACE inhibitors with controls,44 45 46 47 48 49 50 nine compared ARBs with controls,51 52 53 54 55 56 57 58 59 and two compared ACE inhibitors with ARBs.60 61 Seven trials reported specific data for serious pneumonia, five for fatal pneumonia, and eight reported pneumonia without specifying the severity of disease. In only two trials was pneumonia a prespecified outcome.49 59
Among observational studies, 10 were carried out in Asia, five in the United States, and four in Europe. Eleven studies were retrospective and eight were prospective. Seventeen evaluated ACE inhibitors, two ARBs, and two compared ACE inhibitors with ARBs.
The overall quality of the studies was considered to be good. All the randomised controlled trials, except one,51 met the criteria for random sequence generation and about half specifically reported adequate allocation concealment.46 48 50 54 55 57 58 61 Only two randomised controlled trials56 59 were considered to be at high risk of performance bias. Adequate blinding of outcome assessment45 46 47 48 50 52 53 54 55 56 57 58 59 61 and full description of study withdrawals44 45 46 47 48 49 50 51 52 53 54 55 57 58 60 61 were reported in 78% and 89% of randomised controlled trials, respectively. The highest risk of bias was found for potential reporting bias because only two randomised controlled trials presented results for pneumonia as a prespecified outcome.49 59 Supplementary figures 1 and 2 show the results of the quality appraisal of the randomised controlled trials.
All observational studies were considered to have adequate inclusion and exclusion criteria and provided justification for the cohort. Five studies (26%)62 63 64 66 69 did not clearly stated how the drug use was assessed, and four studies (21%)63 64 66 73 did not provide details about outcome assessment. Eleven studies (58%)21 22 68 70 71 72 74 75 77 79 provided results after adjustment for at least one potential variable confounder. In one study76 it was unclear for which variables the results were adjusted. Few studies (26%) reported results adjusted for multiple confounders, and in seven studies (37%) no type of adjustment was mentioned. Supplementary figures 3 and 4 show the results for the quality of the observational studies.
Primary outcome: incidence of pneumonia
Primary outcome data were available from 19 studies comparing ACE inhibitors with controls (five randomised controlled trials, eight cohort or nested case-control studies, and six case-control studies), 11 studies comparing ARBs with controls (nine randomised controlled trials and two cohort or nested case-control studies), and two studies comparing ACE inhibitors with ARBs (one randomised controlled trial and one cohort study).
Use of ACE inhibitors was associated with a significant 34% reduction in risk of pneumonia compared with controls (odds ratio 0.66, 95% confidence interval 0.55 to 0.80; I2=79%). The NNT for 2.0 years was 65 (48 to 112). The magnitude of the risk reduction was similar across all study designs (P=0.78 for subgroup differences). The odds ratios for randomised controlled trials, cohort or nested case-control studies, and case-control studies were 0.69 (0.56 to 0.85; I2=0%), 0.58 (0.38 to 0.88; I2=79%), and 0.67 (0.49 to 0.93; I2=73%), respectively (fig 2⇓).
The risk of pneumonia was not, however, different between patients who did or did not use ARBs (0.95, 0.87 to 1.04; I2=14%). Odds ratio estimates for randomised controlled trials (0.90, 0.79 to 1.01; I2=7%) and cohort or nested case-control studies (1.01, 0.94 to 1.09; I2=0%) did not differ significantly (P=0.10; fig 3⇓).
Pooled results from the two head to head studies showed a non-significant 37% reduction in risk of pneumonia associated with use of ACE inhibitors (0.63, 0.28 to 1.44; I2=78%). In this case, estimates from the randomised controlled trial and cohort study differed significantly (P=0.03) (see supplementary figure 5).
Indirect comparison of ACE inhibitors with ARBs showed a significant 30% reduction in risk of pneumonia associated with use of ACE inhibitors (0.70, 0.56 to 0.86). Similar results were obtained from pooled direct and indirect estimates (0.69, 0.56 to 0.85) without discrepancy (P=0.82) or heterogeneity (I2=0%) between both estimates (fig 4⇓). The NNT for 2.2 years based on this estimate was 72 (51 to 147).
Subgroup analyses for primary outcome
Patients with previous stroke
In patients with previous stroke, use of ACE inhibitors was associated with a 54% reduction in risk of pneumonia compared with controls (0.46, 0.34 to 0.62, I2=0%; seven studies pooled) (see supplementary figure 6). In the same population, however, use of ARBs was not associated with a significant reduction in risk (0.86, 0.67 to 1.09; I2=0%; two studies pooled) (see supplementary figure 7).
The pooled estimate from indirect (odds ratio 0.53, 95% confidence interval 0.16 to 1.79) and direct (0.38, 0.17 to 0.81) evidence of ACE inhibitors compared with ARBs showed a significant 58% reduction in risk of pneumonia (0.42, 0.22 to 0.80; fig 4), without discrepancy (P=0.44) or heterogeneity (I2=0%) between indirect and direct estimates.
Patients with heart failure
In patients with heart failure, two studies evaluated the risk of pneumonia in those treated with ACE inhibitors44 45 and two other studies reported data for those treated with ARBs.51 52 ACE inhibitors were associated with a significant 37% reduction in risk of pneumonia (0.63, 0.47 to 0.84; I2=0%), whereas ARBs showed no significant effect (0.85, 0.49 to 1.47; I2=15%) (see supplementary figure 8).
Patients with chronic kidney disease
In patients with chronic kidney disease, the results from one randomised controlled trial of ACE inhibitors46 (odds ratio 0.15, 95% confidence interval 0.00 to 7.70) and two randomised controlled trials of ARBs52 53 (1.21, 0.32 to 4.52; I2=77%) did not differ significantly when compared with controls (see supplementary figure 9).
Asian and non-Asian patients
Eleven studies were carried out in Asian countries and 11 were done outside of Asia. The PROGRESS20 study was the only multicentre study carried out worldwide that supplied separate data for Asian and non-Asian patients. To lower analysis bias, the other studies carried out worldwide that did not provide separate data for both groups were excluded.
The reduction in risk of pneumonia associated with ACE inhibitors was significantly higher among Asian patients (0.43, 0.34 to 0.54; I2=0%) compared with non-Asian patients (0.82, 0.67 to 1.00, I2=80%, P<0.001 for subgroup differences) (see supplementary figure 10). ARBs, however, were not associated with a reduction in risk of pneumonia in Asian patients (1.04, 0.59 to 1.84; one randomised controlled trial HIJ-CREATE59) or non-Asian patients (0.97, 0.84 to 1.12; I2=27%; five studies pooled; fig 4 and supplementary figure 11).
Secondary outcome: pneumonia related mortality
Data for secondary outcomes were extracted from seven studies comparing ACE inhibitors with controls (three randomised controlled trials and four cohort studies),20 49 50 68 70 71 72 one randomised controlled trial comparing ARBs with control,55 and one head to head randomised controlled trial.60 Five studies comparing ACE inhibitors with controls were carried out on an enriched population—that is, enrolled patients with pneumonia.49 68 70 71 72
Treatment with ACE inhibitors was associated with a significant 27% reduction in risk of pneumonia related mortality compared with controls (0.73, 0.58 to 0.92; I2=51%), without significant differences between estimates from randomised controlled trials and observational studies (P=0.76). The pooled result from randomised controlled trials, however, failed to reach statistical significance (0.61, 0.20 to 1.90; I2=61%) (fig 5⇓).
Only one randomised controlled trial55 reported the effect of treatment with ARBs on pneumonia related mortality (odds ratio 0.63, 95% confidence interval 0.40 to 1.00) (fig 5).
The risk of pneumonia related mortality in indirect (1.16, 0.69 to 1.94), direct (HEAVEN randomised controlled trial60 7.29, 0.14 to 367.24), and pooled comparisons (1.19, 0.71 to 1.98) did not differ between ACE inhibitors and ARBs (fig 4). There was no discrepancy (P=0.36) or heterogeneity (I2=0%) between indirect and direct estimates.
Visual inspection of funnel plots did not reveal any obvious asymmetrical tail (see supplementary figure 12). Publication bias was not suggested by sensitivity analysis taking into account published and unpublished trials (see supplementary figure 13).
In this systematic review we found that treatment with angiotensin converting enzyme (ACE) inhibitors was associated with a significant reduction in risk of pneumonia compared with control treatment and angiotensin receptor blockers (ARBs); the magnitude of this reduction (about one third) was similar across studies with different designs (randomised controlled trials, cohort, and case-control studies). The risk of pneumonia was also reduced in patients treated with ACE inhibitors who were at higher risk of pneumonia, in particular those with stroke and heart failure. Most of the potential protective benefit from ACE inhibitors seemed to be in Asian patients; it is unclear whether the methodology of the studies or the clinical and genetic characteristics of the patients were responsible for this finding. Use of ACE inhibitors was also associated with a reduction in pneumonia related mortality, although the results were less robust than for overall risk of pneumonia; it is uncertain if differences exist between ACE inhibitors and ARBs for this outcome.
The present review was designed to determine the effect of treatment with ACE inhibitors and ARBs on risk of pneumonia. We combined data from both experimental and observational studies to obtain more robust results, mainly because no randomised controlled trial was primarily designed with this objective. Pneumonia is not a rare outcome (particularly in populations treated with ACE inhibitors or ARBs) or an outcome that only occurs months to years after use of ACE inhibitors or ARBs. Therefore randomised controlled trials would have been an appropriate study design to deal with this problem. We found significant statistical heterogeneity for ACE inhibitors but not for ARB results. This was due to the results of observational studies (no heterogeneity was found among randomised controlled trials). Nevertheless, the observed statistical heterogeneity was more quantitative than qualitative because all estimates for study designs share the same direction. This consistency, as well as the robustness of reduction in the risk of pneumonia across all study designs, suggests that use of ACE inhibitors deserves attention. Furthermore, that ACE inhibitors reduced the risk of pneumonia compared not only with the control group but also with ARB treatment, is reassuring because patients’ characteristics and risk factors, as well as other potential clinical and methodological confounders are probably similar between studies on ARBs and those on ACE inhibitors. We were also conservative in our analysis because we did not consider undefined data or data on upper respiratory tract infections, and when studies presented different estimates according to the severity of pneumonia we extracted those reporting only the most severe cases and the most precise or adjusted measure.
Our findings have potential clinical implications. ACE inhibitors are widely prescribed and prescriptions may be influenced by concerns about potential adverse effects, in particular cough, which may be protective. The incidence of ACE inhibitor induced cough has been reported to be in the range of 5% to 35%.80 Our results suggest that patients taking ACE inhibitors who develop cough should, providing that cough is tolerable, persist with treatment. Compliance and persistence with treatment is important. Furthermore, from an evidence based perspective, there is little to choose between ACE inhibitors and the more expensive ARBs. However, in the case of a particular patient, in whom ACE inhibitors and ARBs are presumed to have similar clinical benefit, our results may also influence the choice of prescription in those at high risk of pneumonia. Therefore patients with risk factors for pneumonia and morbidities that require treatment with ACE inhibitors may have an additional reason to continue treatment.
A further important aspect of our results was the reduction in risk of pneumonia across high risk patients, which provided consistency to the overall results. Patients with previous stroke have increased susceptibility to pneumonia owing to risk of aspiration associated with decreased protective reflexes of the respiratory system mediated by substance P and post-stroke dysphagia.14 81 About 20% of these patients will develop pneumonia,82 which is a predictor of poor functional outcome83 84 and a relevant cause of death.84 85 The putative protective effect of ACE inhibitors in this population was predictable given the importance of dysphagia and substance P in these patients. According to one study, ARBs do not increase the levels of substance P or improve asymptomatic dysphagia.86 This highlights the importance of using ACE inhibitors in patients with previous stroke who have comorbidities for which ACE inhibitors are recommended.
Only a few studies evaluated other populations with increased risk, such as patients with heart failure or chronic kidney disease. For patients with heart failure, the decreased risk of pneumonia was also found in patients treated with ACE inhibitors. The suggested effect was significant but this evaluation lacked robust data. ARBs did not show any protective effect.
The putative preventive effect of ACE inhibitors on pneumonia in Asian patients has been suggested.37 We explored this subgroup and compared the effect with non-Asian patients. Furthermore, we obtained a considerable weight of evidence from studies that evaluated Asian patients. ACE inhibitors significantly reduced the risk of pneumonia in both Asian and non-Asian patients, although the odds reduction was significantly higher in Asian patients (57% v 12%; P<0.001). ARBs did not reduce the risk of pneumonia in either population.
Genetic differences in ACE polymorphisms between Asian and non-Asian patients have been suggested to explain the difference in protective effects. Polymorphisms I/I and I/D, which are more prevalent in Asian population, showed a protective trend in the post-hoc analysis in PROGRESS, whereas the D/D polymorphism was less protective.20 77 87 This last polymorphism is associated with acute respiratory distress syndrome, particularly in white populations.88 This potential loss of protective effect may be explained by increased levels of serum ACE inhibitors and catabolism of kinins in patients with the D/D polymorphism.89 However, genetic evidence is equivocal. One study did not find an association between any specific genotype and pneumonia.90 Other factors should be explored to explain these differences in ethnic groups or by geographical location to better define those who can benefit more.
Our conclusions are weaker for pneumonia related mortality because fewer studies provided data for this outcome and significant heterogeneity existed for the results of ACE inhibitors. This uncertainty was reflected by the wider confidence intervals. Treatment with ACE inhibitors (three randomised controlled trials and four cohort studies) and ARBs (one randomised controlled trial) were both associated with a decreased risk of pneumonia related mortality. Explanations for such findings may rely on modulation of cardiovascular risk by ACE inhibitors and ARBs because deaths due to cardiovascular disease are not uncommon among patients with pneumonia.27 91 Decreased mortality may also be explained by the role of ACE inhibitors in pulmonary injury and production of cytokines, which may be related to severity of pneumonia.92 93 94 ACE inhibitors may influence the pattern for release of cytokines exerting anti-inflammatory effects that could reduce the severity of and mortality from pneumonia.95
The influence of ACE inhibitors on survival in these patients should be interpreted carefully because observational studies with enriched populations accounted for most of the weight of the pooled analysis, whereas meta-analysis of three randomised controlled trials (one with an enriched population) did not show differences between ACE inhibitors and controls. However, there was no significant difference in effects between overall randomised controlled trials and observational studies. Although the data were not robust, they did suggest that the effects of treatment with ACE inhibitors on mortality were mostly noticeable in patients with pneumonia.
Limitations of the review
The results and conclusion of this review are weakened by limitations inherent to meta-analysis and individual studies. The overall quality of included studies was good. However, reporting quality for a few studies, particularly observational ones, was low as some of these were abstracted from character limited sections such as letters or comments.
The higher risk of bias was found for potential selective reporting in randomised controlled trials and presentation of unadjusted risk estimates in observational studies. Both limit the strength of our conclusions. A key limitation is that not one randomised controlled trial was primarily designed to assess the effects of ACE inhibitors or ARBs on pneumonia. Although we searched a large number of studies, only a few reported this outcome. Among these, only two randomised controlled trials (<25%) had pneumonia or pneumonia related mortality as a predefined outcome.49 59 As a consequence we were able to extract data only from studies where authors considered pneumonia to be an important outcome, because of either scientific interest or statistical significance.
Observational studies had an important weight in the results for the primary outcome and this should be taken into account when interpreting the clinical implications of our findings. Use of cardiovascular drugs in observational studies could bias results, because patients using drugs could be more concerned for their health and more willing to follow medical advice than controls, the so-called healthy user effect bias.96 However, patients with pneumonia are likely to have a higher risk of cardiovascular disease91 97 and are more likely to be treated with ACE inhibitors, counterbalancing the bias from a healthy user effect. Additionally, the magnitude of the odds risk reduction was similar for randomised controlled trials, cohort studies, and case-control studies.
Pooling data from studies with different designs (confounding bias in observational studies) that evaluated patients in different settings (community based and hospital based studies; referral bias), as well as with different baseline morbidities and heterogeneous risk (membership bias) for pneumonia, should also be taken into account as limitations to our conclusions. The degree of statistical heterogeneity was in fact high in some comparisons. Nevertheless, the pooled estimates from experimental and observational studies were similar. In this case, pooling experimental and observational data increased the power and external validity of the findings.
Included studies compared different ACE inhibitors and ARBs with different controls, such as placebo, calcium channel blockers, and β blockers. In the present analysis we did not carry out serial subgroup analysis to explore if the effect was different for a particular drug because of the scarcity of the data and the risk of obtaining a result by chance.
Finally, we used adjusted indirect comparisons to estimate the effect of ACE inhibitors compared with ARBs. Although combined indirect and direct evidence showed no discrepancies or heterogeneity, the results should not be thought as definitive conclusions because of the possibility of imbalanced data from studies with different designs, baseline risk of patients, and length of follow-up.
Our results suggest an important role of ACE inhibitors, but not ARBs, in reducing the risk of pneumonia. These data may discourage the withdrawal of ACE inhibitors in some patients with tolerable adverse events (namely, cough) who are at particularly high risk of pneumonia. Specific designed randomised controlled trials are required to establish definite conclusions and to estimate better the true magnitude of this putative protective effect. Patients with previous stroke and Asian patients are patient populations that could benefit more from treatment with ACE inhibitors. ACE inhibitors also lowered the risk of pneumonia related mortality, mainly in patients with established disease, but the robustness of the evidence was weaker.
What is already known on this topic
Angiotensin converting enzyme (ACE) inhibitors reduce morbidity and mortality in patients with cardiovascular disease
These drugs also have secondary effects on the respiratory system, suggested to protect against pneumonia
Most of the data on this issue are provided by heterogeneous observational studies with inconclusive results
What this study adds
In pooled results from both interventional and observational studies, ACE inhibitors, but not angiotensin receptor blockers (ARBs), showed a statistical and putative clinically significant protective role against pneumonia
This result may discourage the withdrawal of ACE inhibitors in patients with tolerable adverse events—namely, cough
This protective effect was higher among Asian patients and in those with previous stroke; patient populations that may benefit most from ACE inhibitors
Cite this as: BMJ 2012;345:e4260
We thank the Cochrane Coordinating Centre in Portugal.
Contributors: DC and JA contributed to the concept and design, data acquisition, data analysis, and interpretation of the data; wrote the first draft of the manuscript; critically revised the manuscript; and gave final approval of the submitted manuscript. AVC contributed to the interpretation of data, critically revised the manuscript, and gave final approval of the submitted manuscript. JC contributed to the concept and design, data analysis, and interpretation of the data; wrote the first draft of the manuscript; critically revised the manuscript; and gave final approval of the submitted manuscript. JC is the guarantor.
Funding: This was an academic project not funded by government or non-government grants.
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 organisation for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work 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.