Microbiological Etiologies of Pneumonia Complicating Stroke
A Systematic Review
Background and Purpose—Identifying the causal pathogens of pneumonia complicating stroke is challenging, and antibiotics used are often broad spectrum, without recourse to the microbiological cause. We aimed to review existing literature to identify organisms responsible for pneumonia complicating stroke, before developing a consensus-based approach to antibiotic treatment.
Methods—A systematic literature review of multiple electronic databases using predefined search criteria was undertaken, in accordance with Cochrane and PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidance. Published studies of hospitalized adults with ischemic stroke, intracerebral hemorrhage, or both, which identified microbiological etiologies for pneumonia complicating stroke up to January 1, 2017, were considered. Analysis included summary statistics and random-effects meta-analysis where appropriate.
Results—Fifteen studies (40% ischemic stroke, 60% ischemic stroke and intracerebral hemorrhage) involving 7968 patients were included. Reported occurrence of pneumonia varied considerably between studies (2%–63%) with a pooled frequency of 23% (95% confidence interval, 14%–34%; I2=99%). Where reported (60%), the majority of pneumonia occurred within 1 week of stroke (78%). Reported frequency of positive culture data (15%–88%) varied widely. When isolated, aerobic Gram-negative bacilli (38%) and Gram-positive cocci (16%) were most frequently cultured; commonly isolated organisms included Enterobacteriaceae (21.8%: Klebsiella pneumoniae, 12.8% and Escherichia coli, 9%), Staphylococcus aureus (10.1%), Pseudomonas aeruginosa (6%), Acinetobacter baumanii (4.6%), and Streptococcus pneumoniae (3.5%). Sputum was most commonly used to identify pathogens, in isolation (40%) or in conjunction with tracheal aspirate (15%) or blood culture (20%).
Conclusions—Although the analysis was limited by small and heterogeneous study populations, limiting determination of microbiological causality, this review suggests aerobic Gram-negative bacilli and Gram-positive cocci are frequently associated with pneumonia complicating stroke. This supports the need for appropriately designed studies to determine microbial cause and a consensus-based approach in antibiotic usage and further targeted antibiotic treatment trials for enhanced antibiotic stewardship.
Pneumonia complicating stroke occurs frequently, independently increasing mortality 3-fold and increasing hospitalization costs, length of stay, and likelihood of poor outcome in survivors.1,2 Although diagnosis remains challenging, the PISCES (Pneumonia in Stroke Consensus) group recommended that Stroke-Associated Pneumonia (SAP) was the preferred diagnostic terminology covering the spectrum of lower respiratory tract infections complicating stroke within the first week and hospital-acquired pneumonia (HAP) after 1 week.3 Furthermore, acknowledging the limitations of current biomarkers and accessibility of microbiological samples, modified Centers for Disease Control and Prevention criteria were proposed to aid clinicians and researchers in diagnosing SAP in nonventilated patients.3
Once SAP is suspected or diagnosed, however, use of antimicrobials vary and are either clinician dependant or guided by local policy for community-acquired pneumonia (CAP) or HAP.4 Antibiotics used are often broad spectrum, without recourse to the microbiological cause. The ability to better inform choice of antibiotic therapy in SAP, based on defined or likely microbial cause, might lead to improved outcomes and enhanced antibiotic stewardship. Identifying microbiological cause in nonventilated stroke patients is challenging because of the difficulties in obtaining direct samples from the lower respiratory tract (impaired cough and limited expectoration) and lack of applicable invasive procedures, such as bronchoscopy in conscious stroke patients, in addition to reliance on sputum samples with the inherent risk of contamination from oropharyngeal commensal organisms.5 Although bacterial colonization of the oropharynx could potentially limit interpretation of positive sputum samples, poor diagnostic sensitivity of microbiological culture methods, such as blood culture specimens (positive in <10%)6 and pleural fluid aspirate, limit their use. Most importantly, prior use of antibiotics hampers the sensitivity of microbiological techniques, and current stroke guidelines do not recommend early nonselective preventive antibiotic treatment.7,8 As part of the ongoing PISCES collaboration, we sought to identify microbiological etiologies for pneumonia complicating stroke through a systematic review of available literature to help inform a planned consensus-based approach for antibiotic treatment.
A systematic literature review was undertaken in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement and Cochrane guidance.9,10 The authors declare that all supporting data are available within the article (and its online-only Data Supplement).
Data Sources and Searches
Searches were undertaken in medical databases including Medical Literature Analysis and Retrieval System Online (MEDLINE, National Institute of Health Library Interface, 1946–January 1, 2017), Excerpta Medica database (EMBASE, National Institute of Health Library Interface, 1947–January 1, 2017), Cumulative Index of Nursing and Allied Health Literature (CINAHL, National Institute of Health Library interface, 1937–January 1, 2017) and Cochrane Central Register of Controlled Trials (Wiley interface, current issue) using predefined search criteria and terms (Table I in the online-only Data Supplement). Hand searching of reference lists for additional eligible articles was also performed, and the members of PISCES group were invited to provide any other potentially eligible articles. Non-English full-text articles were translated and considered for inclusion if eligibility criteria were met.
Published studies of hospitalized adults with ischemic stroke, intracerebral hemorrhage, or both, which identified any potential pathogen responsible for pneumonia complicating stroke or which had used objective criteria for diagnosing pneumonia complicating stroke (but not reported causative organisms), were independently screened for eligibility by 2 reviewers (A.K.K. and C.J.S.), using the study title and abstract (Table II in the online-only Data Supplement). For those studies that used objective criteria for pneumonia but which had not reported causative pathogens, corresponding authors were contacted by e-mail for unpublished information on pathogens responsible for pneumonia if available. Lead or corresponding authors of studies under consideration were also contacted by e-mail to resolve any issues relating to assessment of eligibility or data extraction. Discrepancies relating to eligibility or data extraction were resolved by a consensus discussion between the same 2 study investigators.
Data were independently extracted by 2 reviewers (A.K.K. and A.V.) and included study design, sample size, publication status and demographical data (year of study, country of study, clinical environment), stroke type (ischemic, intracerebral hemorrhage, or both), interval from admission to diagnosis of pneumonia, frequency of pneumonia, criteria used for pneumonia diagnosis, mean age, mean National Institutes of Health Stroke Scale score, cardiovascular risk profile, swallow screening, proportion of nil oral/tube-fed, type of culture specimen (sputum, blood culture, pleural or tracheal aspirate, serology), organisms identified, and antibiotic usage (prophylactic or treatment). Reported data in the identified eligible publications were supplemented by contacting corresponding or lead authors where necessary.
The primary outcomes were the frequencies of the most commonly isolated microbial species among the included studies and the proportion of these organisms responsible for pneumonia. Secondary outcomes included (1) the frequency of positive microbiological cultures and proportions of isolated organisms in positive cultures; (2) relationships between microbial species and time interval from stroke onset to pneumonia; and (3) frequency of identified pathogens across different geographical regions.
Risk of Bias and Quality Assessment
We anticipated inclusion of both randomized controlled trials and nonrandomized studies that had pneumonia as an outcome but were not primarily designed to identify microbiological etiologies. Hence, a formal statistical tool for assessing bias or the individual quality of the studies was not used as we were less concerned about the design or the effects of interventions used in the individual studies for this review. However, as cultures for organisms would only be undertaken when pneumonia is suspected or diagnosed, heterogeneity among studies reporting pneumonia was assessed using random-effects model. Heterogeneity was quantified with the I2 statistic as reported in a previous study.1 This measures the proportion of variation ascribed to excess heterogeneity beyond that anticipated by chance. Because of anticipated heterogeneity between studies reporting pneumonia, further quantitative meta-analysis on bacterial cause was not undertaken. Summary statistics were instead undertaken to describe primary and secondary outcomes. A post hoc descriptive comparison of the frequency of pathogen species detected in pneumonia complicating stroke with other forms of pneumonia (eg, CAP, HAP) was also undertaken.
A total of 6231 unique publications were identified by electronic searches and through the PISCES collaborators (Figure 1). Fifteen fully published studies were finally considered eligible for inclusion.11–25
Study and Patient Characteristics
The studies included retrospective (40%) or prospective (33%) observational studies and randomized trials (27%). Fifty-four percent were European or North American studies; 46% were performed in South East Asia or Asia Pacific region. The majority of the studies were conducted on the acute stroke unit (72%). Other clinical environments included rehabilitation wards (7.5%) and intensive care units (15%). The mean age of the patients in individual studies ranged from 58 to 83 years. Baseline stroke severity (National Institutes of Health Stroke Scale) was reported in only 73% of studies, with a mean ranging from 5 to 19. Three studies were randomized controlled trials of prophylactic antibiotics.11,15,25 Two studies were exclusively in nasogastric or percutaneous endoscopic gastrostomy tube-fed participants.21,23 Thirty-three percent of the studies included mechanically ventilated patients. Entry criteria to the studies varied widely, with only 1 study having the identification of microbiological cause for pneumonia complicating stroke as a primary objective12; 2 studies excluded patients on immunosuppressive medication, with prior malignancy, or other forms of immunosuppression before stroke.11,22 Other comorbidities and cardiovascular risk factor profiles were reported varyingly (Table III in the online-only Data Supplement; Table 1).
Diagnosis and Frequency of Pneumonia
Reported occurrence of pneumonia varied between studies (2%–63%). Pooled frequency of reported pneumonia was 23% (95% confidence interval, 14%–34%; I2=99%; Figure 2). Substantial heterogeneity was noted even when adjusted to stroke subtype (ischemic stroke, I2=96%; mixed ischemic stroke and intracerebral hemorrhage, I2=99%) or geographical location (Asian studies, I2=97.4% and European or North American studies, I2=97.8%). When reported (60%), the majority of pneumonia occurred within 1 week of stroke (78%). The Centers for Disease Control and Prevention criteria (45%) and ad hoc objective criteria (40%) were the most commonly used objective criteria to diagnose pneumonia.
Sputum culture was most commonly used to identify pathogens either in isolation (40%) or in conjunction with tracheal aspirate (15%) and blood culture (20%). Reported frequency of positive culture data (15%–88%) varied considerably. Only 3 studies described the culture methods used to identify organisms.4,13,25 No bacterial growths were reported in 2 studies (15% and 67%).12,25 Identification of bacterial species in positive cultures varied between studies (Table IV in the online-only Data Supplement). No pathogen was identified in every study although Staphylococcus aureus was identified in positive cultures in 14 of 15 (93%) studies, whereas Acinetobacter baumanii was identified in positive cultures in only 6 of 15 (40%) studies (Table IV in the online-only Data Supplement; Figure 3). Antibiotic susceptibility was not reported in the majority of studies; no studies reported any viral or other atypical organisms although it was unclear if these were tested for.
The proportions of microbial species associated with pneumonia also varied between studies. Overall, aerobic Gram-negative bacilli (38%) and Gram-positive cocci (16%) were most frequently responsible for pneumonia; commonly isolated phenotypes (Table 2) included Enterobacteriaceae (21.8%: Klebsiella pneumoniae, 12.8% and Escherichia coli, 9%), S. aureus (10.1%), Pseudomonas aeruginosa (6%), A. baumanii (4.6%), and Streptococcus pneumoniae (3.5%). Studies that included patients at relatively higher-risk of pneumonia, that is, exclusively dysphagic patients or intensive care studies,15,17,19,21 were found to have a high proportion of aerobic Gram-negative bacilli and S. aureus, in comparison to lower risk studies (unselected stroke patients with ≤15% intracerebral hemorrhage and mean National Institutes of Health Stroke Scale score ≤5; Table 1). It was not possible to explore relationships between timing of pneumonia or its severity with individual organisms because of insufficient data.
We compared the frequencies of the 8 most commonly identified organisms in pneumonia complicating stroke with those of hospitalized CAP, ventilator-associated pneumonia (VAP), and HAP (Table 2) from recent reviews of literature (terminologies defined in Table V in the online-only Data Supplement).26–29 Geographical variations in bacterial cause were observed in our study as seen as in the other reviews. In particular, Gram-negative opportunistic pathogens, such as P aeruginosa and A baumanii, were more commonly isolated in South Asia or Asia Pacific regions (75% and 100%) as opposed to Western Europe or the United States (28% and 0%). Several organisms were reported with comparable frequency (range) to VAP or HAP (eg, K pneumoniae, E coli). Spneumoniae, the organism most frequently identified in CAP, was detected less often in pneumonia complicating stroke. The organisms most often reported in HAP and VAP (S aureus and P aeruginosa) were also identified less frequently in pneumonia complicating stroke.
Only 4 studies (24%) identified antibiotics used to treat pneumonia complicating stroke.13,15,20,25 The antibiotic of choice was determined by local hospital policy and commonly included β-lactam (including ureidopenicillin and second/third generation cephalosporins) antibiotics±β-lactamase inhibitors and second/third generation fluoroquinolones and was always initiated before obtaining antibiotic sensitivities. Only 1 study reported the proportion of patients with pneumonia receiving antibiotics, the number of pneumonia episodes, and functional outcomes after treatment with antibiotics.25
Pneumonia occurs most frequently during the first week after stroke (SAP)1,3 and may, therefore, include microbiological etiologies associated with hospitalized CAP or HAP. Our study suggests that aerobic Gram-negative bacilli (eg, K pneumoniae, E coli, and P aeruginosa) and Gram-positive cocci (eg, S aureus and S pneumoniae) were associated with the majority of pneumonia complicating stroke when cultures were sent. A recent review suggested that ≈80% of hospitalized CAP were caused by S pneumoniae, Mycoplasma pneumoniae, Chlamydia pneumoniae, and Haemophilus influenza.26 The same review found that ≈50% of HAP was caused by S aureus and 35% by Pseudomonas species, Klebsiella species, Escherichia species, Acinetobacter species, and Enterobacter species.26 The spectrum of identified organisms in our study seems to be more closely related to HAP than VAP or hospitalized CAP although antibiotic susceptibilities were not reported in most of the included studies. Furthermore, none of the included studies reported the results of investigations for viral or atypical pathogens, or even if they were done. Although our study is a comprehensive systematic literature research and collaboration within the PISCES group, our findings need to be interpreted with caution because of several inherent limitations.
First, there was marked heterogeneity between the included studies, which likely contributed to the variation in identified organisms and their relative contributions to pneumonia. For example, studies undertaken in the critical care environment might yield a higher proportion of organisms overall because of access to more direct sampling (eg, bronchoscopy or tracheal aspirate) and also more frequent causes because of organisms typical of HAP or VAP.
We could not identify any single factor that fully explained the high heterogeneity (I2=99%). Although both stroke type and our prospective risk categorization (Table 2) showed anticipated differences in pneumonia frequency, even within our very high risk category frequency ranged from 14% to 63% (Figure 2), and heterogeneity remained extremely high (I2>90%) within each risk category. The asymmetry of the scatterplots seen in the funnel plot also reflects the high heterogeneity among studies, which was not corrected even when subgroup analysis was undertaken with pneumonia risk stratification (Figure I in the online-only Data Supplement). Sixty-six percent of the included studies in our review were deemed high or very high risk (Table I in the online-only Data Supplement) for developing pneumonia reflecting the higher frequency of pneumonia seen in this review as opposed to a previous systematic review.1 Apart from patient selection, the overall high heterogeneity also reflects varying geographical location, different study designs, inclusion criteria, timing from stroke onset to sampling, and differences in standardized outcome definition for SAP. We were unable to differentiate pneumonia and causative pathogens for patients admitted from institutional environments, such as nursing homes, which may have also contributed to heterogeneity. Stroke registries, not routinely expected to collect and maintain data on microbiological cause, were also excluded unless specific mention was made about determining bacterial cause, which could have contributed to selection bias and heterogeneity. The variation in approach to diagnosis of pneumonia complicating stroke is well recognized,1 and may, therefore, influence the threshold for sending microbiological samples, contributing to verification bias. Second, only one of the studies was primarily designed to identify the microbial cause of pneumonia complicating stroke,12 whereas the remaining studies collected microbiological data when available within the context of their individual study objectives. It is reasonable to assume that the proportion of positive cultures could be higher if culture samples were sent systematically in all suspected cases of pneumonia. However, compared with other clinical settings (eg, CAP or VAP), consistently obtaining sputum samples in nonventilated stroke patients is challenging and alternative strategies (eg, bronchoscopy for VAP) are limited. Third, it was unclear among most studies as to when culture samples were sent in relation to onset of stroke and suspicion or diagnosis of pneumonia. Although the majority of pneumonia in our review occurred within a week of stroke symptom onset, it was not possible to further explore microbiological etiologies in relation to timing of pneumonia relative to stroke symptom onset. Although one could hypothesize that organisms commonly associated with CAP are most likely causal in early SAP (≤72 hours), and those associated with HAP causal in SAP beyond 72 hours, we were unable to confirm or refute this finding because of limited data in the individual studies. Our observations of an apparent low yield of CAP organisms, and higher yield of HAP organisms could be at least in part because of sampling bias beyond 48 to 72 hours after stroke onset. Fourth, microbiological methods used to collect sputum samples, number of specimens sent when pneumonia was diagnosed, delays in analyzing samples if any, and laboratory techniques used were inadequately reported. None of the studies used modern molecular-based polymerase chain reaction methods or urinary antigen testing (for organisms, such as S. pneumoniae and Legionella pneumophilia). This may also contribute to differences in the observed frequencies of positive culture data and the apparent lack of atypical or viral causes. Interestingly, a recent study reviewing hospitalized patients identified ≈22% of CAP inpatients had viral pathogens (most commonly rhinovirus, 9% and influenza, 6%) implying that a viral cause of SAP in at least some individuals may be possible.30 Finally, antibiotics preceding index stroke (especially for patients with chronic lung disease) may have influenced microbiological cultures. For example, in the PANTHERIS study (Preventive Antibacterial Therapy in Acute Ischemic Stroke), when sputum samples were analyzed, 36% of samples were positive for organisms in the placebo group as opposed to 9% in the prophylactic antibiotic group.15 Although the numbers are too small to form further conclusions, prophylactic antibiotics administered to the participating patients with stroke in the 3 randomized controlled trials may have affected frequency of identified organisms.
It is important to emphasize the differences in frequency of pathogens identified in Asian compared with European or North American studies. For example, in a study of hospitalized CAP patients, S. pneumoniae appeared to have a lower frequency (13% versus 26%) and Enterobacteriaceae appeared to have a higher frequency (9% versus 2.7%) in Asian studies as compared with European studies.5 Similarly, although limited comparison was possible in our study, the frequency of certain nosocomial pathogens appeared to be higher in Asia or Asia Pacific regions (Table IV in the online-only Data Supplement) in keeping with higher prevalence of hospital-acquired infection (15.5%) in comparison to Europe (7.1%) or the United States (4%).31 However, this incongruity may be as a result of selection pressure on clinically relevant bacteria from differing prior antibiotic exposure of patients across continents, as well as possible implementation of pneumococcal vaccination programmes.
The scarce amount of available data on microbiological etiologies of pneumonia complicating stroke might also reflect the clinical routine, as suggested by a recent survey on German stroke units.4 Treatment guidelines for pneumonia complicating stroke, nevertheless, should take into account these commonly isolated organisms and also consider local surveillance data, community pathogens, and demographical variations, together with guidance from the World Health Organization global strategy for containment of antimicrobial resistance32 when recommendations are being made for prescribing empirical antibiotic regimens. Anaerobes, although not identified in our study, are commonly seen in the upper airway mixed with oral flora and in the stomach and are often thought to be responsible for aspiration pneumonia. However, anaerobes are difficult to culture, and if aspiration is suspected, then broader spectrum antibiotics may be required.28
The risk of contamination with oral flora, low diagnostic yield (30%–40% sensitivity) with current diagnostic methods,31 and delay in producing a positive result (at least 24–48 hours) often predisposes to initial broad-spectrum antibiotic prescriptions. An ideal diagnostic method would be more timely and sensitive to identifying pathogens. Polymerase chain reaction assays involving comprehensive molecular testing platforms significantly improve pathogen detection (87% versus 39%) in comparison to sputum culture (including viral pathogens) and provide results within 24 hours, which may help in initiation of pathogen-directed microbial therapy or a rapid de-escalation of broad-spectrum antibiotic therapy.29 However, validating such technology still depends on a reliable microbiological reference standard (sputum analysis), which may limit its potential utility in nonventilated stroke patients.
Our study demonstrates an evidence gap in appropriately designed studies that robustly identify microbiological etiologies in pneumonia complicating stroke. Although limited by small and heterogeneous study samples, this review suggests aerobic Gram-negative bacilli and Gram-positive cocci species are frequently associated with pneumonia complicating stroke. Difficulties in obtaining suitable sputum samples among nonventilated stroke patients and poor sensitivity of current diagnostic methods often result in broad-spectrum antibiotic prescriptions for pneumonia. Our study, however, supports the need for a consensus-based approach to antibiotic initiation and further targeted antibiotic treatment trials for enhanced antibiotic stewardship.
We acknowledge Dr Ryan Keh (Speciality Trainee in Neurology, Salford Royal NHS Foundation Trust) and Dr Luciana Miguel Alhambra (Clinical Fellow in Stroke Medicine, Salford Royal NHS Foundation Trust) for their assistance in translating non-English language studies for this review.
Sources of Funding
Dr Meisel is supported by the German Research Foundation (EXC257 and SFB-TR84).
Dr Meisel received project funding by Thermo Fisher Scientific BRAHMS GmbH, Germany for a stroke trial. The other authors report no conflicts.
Guest Editor for this article was Harold P. Adams, MD.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.117.020250/-/DC1.
- Received November 28, 2017.
- Revision received May 2, 2018.
- Accepted May 14, 2018.
- © 2018 American Heart Association, Inc.
- Kishore AK,
- Vail A,
- Chamorro A,
- Garau J,
- Hopkins SJ,
- Di Napoli M,
- et al
- Smith CJ,
- Kishore AK,
- Vail A,
- Chamorro A,
- Garau J,
- Hopkins SJ,
- et al
- Harms H,
- Hoffmann S,
- Malzahn U,
- Ohlraun S,
- Heuschmann P,
- Meisel A
- Adams HP Jr,
- del Zoppo G,
- Alberts MJ,
- Bhatt DL,
- Brass L,
- Furlan A,
- et al
- Shamseer L,
- Moher D,
- Clarke M,
- Ghersi D,
- Liberati A,
- Petticrew M,
- et al
- Higgins JPT,
- Green S
- Chamorro A,
- Horcajada JP,
- Obach V,
- Vargas M,
- Revilla M,
- Torres F,
- et al
- Vargas M,
- Horcajada JP,
- Obach V,
- Revilla M,
- Cervera A,
- Torres F,
- et al
- Becker KJ,
- Dankwa D,
- Lee R,
- Schulze J,
- Zierath D,
- Tanzi P,
- et al
- Warusevitane A,
- Karunatilake D,
- Sim J,
- Lally F,
- Roffe C
- 32.↵WHO (World Health Organization). Global Action Plan on Antimicrobial Resistance. http://www.who.int/antimicrobial-resistance/publications/global-action-plan/en/. Accessed November 10, 2017.