Advances in the diagnosis of pneumonia in childrenBMJ 2017; 358 doi: https://doi.org/10.1136/bmj.j2739 (Published 26 July 2017) Cite this as: BMJ 2017;358:j2739
- Heather J Zar, professor and chair, Department of Paediatrics and Child Health1,
- Savvas Andronikou, professor1 2,
- Mark P Nicol, professor and head, Division of Medical Microbiology3 4
- 1Department of Paediatrics and Child Health, Red Cross Children’s Hospital and MRC Unit on Child and Adolescent Health, University of Cape Town, Cape Town 7700, South Africa
- 2Department of Radiology, University of Bristol and the Bristol Royal Hospital for Children, UK
- 3Division of Medical Microbiology, Department of Pathology, University of Cape Town
- 4National Health Laboratory Service of South Africa, Johannesburg, South Africa
- Correspondence to:
H J Zar
Pneumonia remains a major cause of childhood mortality and morbidity globally. Accurate diagnosis and attribution of the causes of pneumonia are important for measuring the burden of disease, implementing appropriate preventive or treatment strategies, and developing more effective interventions. This review summarizes recent diagnostic advances in radiological techniques, specimen collection, and laboratory methods. Although chest ultrasound and chest magnetic resonance imaging are promising modalities for radiological diagnosis, their role in clinical management and their impact on outcomes need further study. Rapid, highly sensitive, multiplex laboratory tests performed on upper respiratory tract samples or induced sputum can detect nucleic acid from potential pathogens in most children with pneumonia. However, it may be difficult to attribute causality because it is often impossible to distinguish between organisms colonizing or infecting the upper respiratory tract and those causing pneumonia. Currently available host biomarkers lack accuracy for distinguishing bacterial or mixed bacterial-viral infections from viral infections. New biomarkers derived from host transcriptional profile analysis may be more accurate but require validation. Prospective studies with appropriate control populations, including studies of clinical impact, are needed to improve our understanding of the role of tests. Although progress has been made in radiological techniques and laboratory testing, current methods for diagnosing and attributing the causes of pneumonia are suboptimal.
The incidence and severity of pneumonia in childhood, and its related mortality, have declined substantially in the past decade owing to improved socioeconomic conditions; better access to care; wider implementation of effective management and preventive strategies; and the development and availability of improved vaccines, particularly the pneumococcal (PCV) and Haemophilus influenzae type b (Hib) conjugate vaccines.1 However, pneumonia remains the leading cause of childhood mortality globally outside the neonatal period and a major cause of morbidity and hospital admission.23 Furthermore, pneumonia or lower respiratory tract infection (LRTI) in early childhood has increasingly been associated with reduced lung function in infancy and the development of chronic non-communicable respiratory diseases, such as asthma or chronic obstructive pulmonary disease, in children and adults.4567
Pneumonia was formerly thought to occur after invasion of a sterile lower respiratory tract by a single pathogenic organism. However, recent evidence indicates that the healthy lung is not sterile and that the normal lung microbiome exists in a dynamic state.8 A key step in the pathogenesis of pneumonia may be dysbiosis, or imbalance, in the normal microbial ecology of the respiratory tract as a result of factors related to the host, environment, or organism.910 Dysbiosis may result in the overgrowth of a single pathogen, or multiple pathogens, as supported by recent data from studies using sensitive highly multiplexed diagnostics, which identify multiple potentially pathogenic organisms from the respiratory tract of most children with pneumonia.1112
Accurate diagnosis of pneumonia and the attribution of its cause are important for measuring the burden of disease, for implementing appropriate treatment or preventive strategies, and for developing new more effective interventions. The availability and increased uptake of global and national childhood immunization programs of conjugate vaccines against bacterial pathogens have had an impact on the causes of childhood pneumonia. The relative contribution of different pathogens needs to be re-evaluated to inform the development of modified treatment guidelines, such as the widely used World Health Organization Integrated Management of Childhood Illness guideline, and to prioritize vaccine development and implementation.
Diagnostic advances include the use of new radiological methods, better specimen collection, and improved microbiological tests. Advances in radiological techniques include the use of point-of-care (POC) chest ultrasound and chest magnetic resonance imaging (MRI). In addition, telemedicine enables the remote interpretation of radiological investigations by expert radiologists. Improved methods for specimen collection include a better understanding of the role of non-invasive specimens (such as urine or nasopharyngeal samples) and the use of induced sputum. Laboratory based advances encompass rapid tests for pathogen nucleic acid or antigen, simultaneous detection of multiple pathogens, and host transcriptional profile analysis for the identification of novel biomarkers.
This review summarizes recent advances in the diagnosis of community acquired pneumonia in childhood and focuses on radiological techniques, specimen collection, and laboratory methods. The impact of these advances on the current understanding of the pathogenesis of pneumonia and the implications for future strategies and research are discussed.
Incidence of pneumonia in childhood
The incidence of pneumonia in childhood has fallen substantially in recent decades, but pneumonia remains the main cause of death in children outside the neonatal period and of loss of disability adjusted life years in children and adolescents.23 The Global Burden of Diseases study, which used many data sources and mathematical modeling, estimated that pneumonia accounted for almost 900 000 of the 6.3 million deaths in children in 2013, with the greatest burden occurring in low and middle income countries.2 The incidence of pneumonia in children under 5 years of age in high income countries has been estimated at 0.015 episodes per child year, compared with 0.22 episodes per child year in low and middle income countries.13
Several case-control or descriptive studies, mainly in high income countries, have reported that vaccination, especially with the 13-valent PCV (PCV13), has reduced the incidence and severity of pneumonia in children and of complications such as empyema.1415161718 Nevertheless, the incidence of pneumonia in childhood, even in highly vaccinated populations, imposes a substantial burden on healthcare systems. The incidence of pneumonia in children participating in a South African birth cohort study in a peri-urban, poor socioeconomic area was 0.27 episodes per child year during infancy, with the peak incidence occurring at 3 months of age19; this occurred despite high coverage of vaccination, including PCV13.
Specific childhood populations may be especially vulnerable to developing pneumonia or severe disease. A meta-analysis reported that the incidence of and mortality from pneumonia in children infected with HIV is about six times higher than in those not infected with HIV.20 Infants who have been exposed to HIV but are not infected (those born to HIV infected mothers who are HIV negative owing to effective interventions for preventing mother-to-child transmission) are also increasingly recognized as being more vulnerable to pneumonia. These infants have a higher risk of severe disease or hospital admission than children who have not been exposed to HIV (infants born to HIV negative mothers).21 A meta-analysis confirmed several other risk factors for pneumonia associated mortality including malnutrition, lack of breast feeding, crowded living conditions, exposure to indoor air pollution, and low birth weight.22
Sources and selection criteria
This review examines advances in the diagnosis of pneumonia in children, focusing on radiological techniques, specimen collection, and laboratory methods. We identified references through searches of publications listed in PubMed from January 2000 to September 2016 using combinations of medical subject headings (MeSH) that included child AND (pneumonia OR lower respiratory tract infection OR pertussis) AND (diagnosis OR chest radiology OR lung ultrasound OR MRI lung OR induced sputum OR specimen). We reviewed relevant titles and abstracts, and we prioritized randomized studies, systematic reviews or meta-analyses, and case-control or cohort studies. Where higher quality evidence was not available, case series and other observational studies were considered. Individual case reports were not included. Preference was given to publications from the past five years. Because this was not a formal systematic review, we did not grade the included studies but summarized key outcomes or results. We only included articles or abstracts published in English.
Current guidelines for radiological diagnosis
Chest radiography is often performed for suspected pneumonia,2324 which is the most common indication for imaging of the chest.25 This is despite several guidelines advising against the routine use of chest radiography25 and the lack of evidence for its impact on clinical outcomes.2627 Several international or national guidelines including those from the British Thoracic Society (BTS)28 and the National Institute for Health and Care Excellence (NICE) in the United Kingdom,29 the Paediatric Infectious Diseases Society and the Infectious Diseases Society of America (IDSA),30 the South African Thoracic Society (SATS),31 and the International Union against Tuberculosis and Lung Diseases32 do not recommend chest radiography in children who are well enough to be treated as outpatients. Rather, chest radiography is recommended only in children who are admitted to hospital with severe symptoms, hypoxia, or suspected complications such as empyema.
Limitations of chest radiography
Changes seen on chest radiographs may be more specific for pneumonia than clinical signs, with radiological signs associated with more severe disease and treatment failure.3334 However, the use of chest radiography for diagnosis has several limitations. The two dimensional nature of chest radiography may lead to consolidation, adenopathy, or complications being masked by other anatomical structures such as the heart, mediastinum, and diaphragm3536; it may also lead to the problem of summation shadows.37 Furthermore, inter-reader agreement in the interpretation of chest radiographs is poor.232538 In a recent review, 10 of 12 pediatric studies showed only fair to moderate interobserver agreement.38 Agreement is worse for reporting of an “infiltrate” in children under 5 years,23 and it remains moderate even when using WHO standardized radiological criteria for pneumonia.2339 In addition, the lack of abnormality on chest radiography does not exclude pneumonia,4041 and abnormal chest radiographs may be interpreted as normal.23 Other limitations include lack of access to basic radiology services at primary healthcare facilities in many low and middle income countries; expense; exposure to radiation (although low at 0.01-0.02 mSv for a standard chest radiograph)42; and need for specialized equipment, power source, and trained technologists. Imaging with computed tomography is more reliable but discouraged for routine diagnosis because of the resources needed and the radiation risk,25 despite new dose reduction techniques.43
New imaging techniques are therefore needed to increase diagnostic accuracy and improve access. Two alternative forms of diagnostic imaging have emerged—POC ultrasound and rapid sequence MRI, which may have applications across different settings. The use of these techniques is enhanced by the capacity for remote reading by expert radiologists.
Advances in radiological diagnosis
Clinician led POC ultrasound
It is now feasible for ultrasound to be performed by non-radiologists with the advent of affordable handheld machines. Children are ideal candidates for diagnostic imaging using ultrasound because of their thin chest wall and small lung mass. Ultrasound has several advantages over other imaging modalities: it can be performed at POC, it is feasible and less costly than chest radiography, it is less affected by movement or crying than other imaging modalities, it can be done in sleeping children,35 and it is free of ionizing radiation.243541 Although ultrasound performed by non-radiologists is controversial, it has become a useful tool for physicians, emergency medicine doctors, and intensivists.44
Lung ultrasound involves scanning a child in the anterior, lateral, and posterior areas of each hemithorax.454647 Consolidation is diagnosed as a hypoechoic area with ill defined borders surrounded by B-lines and loss of or an irregular shredded border of the pleural line.244547484950 Other signs include punctate hyperechoic or linear and branching echogenic structures, reflecting air bronchograms, and a decrease in lung sliding (fig 1⇓). Lobar involvement presents as a liver-like area in the thorax or hepatization of the lung.47 Homogeneous, anechoic, or hypoechoic fluid in the pleural space indicates a pleural effusion (fig 2⇓).47
Evidence for the use of ultrasound to support the diagnosis of pneumonia in children is accumulating. Most studies have compared POC ultrasound performed by clinicians with chest radiography findings reported by radiologists (table 1⇓). A literature search of lung ultrasound for pneumonia in children yielded 24 original studies, 11 commentaries, a meta-analysis, and a summary of evidence.62 Only three studies compared ultrasound with a gold standard of pneumonia defined by computed tomography, and only one study provided the diagnostic accuracy of ultrasound and chest radiography against computed tomography for pneumonia. In this last study the positive predictive values for ultrasound and chest radiography were 0.61 and 0.71, respectively, whereas the negative predictive values were 0.86 and 0.8, respectively.58 Four studies provided interobserver agreement, with three reporting high agreement for consolidation on ultrasound (κ=0.8-0.9).505459 Only one study reported the interobserver agreement for chest radiography and ultrasound simultaneously; agreement for ultrasound was fair (κ=0.55) but better than that for chest radiography (κ=0.33).58
The limitations of ultrasound include its inability to visualize the whole lung at the same time or to identify consolidation deep within the lung parenchyma.25 In addition, the spleen or air in the stomach can be misinterpreted as lung consolidation with air bronchograms.54
Overall, the use of POC ultrasound as an alternative to chest radiography in primary screening for triage in the emergency department or clinic is promising. However, even though ultrasound has potential as a screening test, it is not clear whether it affects patient outcome or management. Further studies are needed to compare the impact of chest radiography versus POC ultrasound on patient relevant outcomes and to inform best practice.
Rapid sequence MRI
When cross sectional imaging is needed for the diagnosis of severe pneumonia or associated complications, the possible radiation risks of computed tomography can be avoided by using MRI.6364 It can be difficult to achieve adequate MRI imaging of the lungs owing to low proton density with signal loss and because cardiac pulsation and breathing produce strong motion artefacts.37636566 Furthermore, in children who are too young to cooperate, sedation or anesthesia is required, and this may cause dependent atelectasis, which can be misinterpreted as infection.6567 Therefore, older children are scanned awake, and in cooperative children respiratory motion can be prevented by voluntary breath holding.65
At some facilities MRI has become an established technique for imaging the lungs using improved sequences and modern respiratory triggering6365—a mechanism of scanning in between breaths. Faster protocols using conventional T2 weighted sequences have been developed that can be performed with free breathing and completed in under 10 minutes, enabling imaging in children with pneumonia.6466
Disease in the alveolar spaces results in a high intensity MRI signal that stands out against the low signal of normally aerated lung (fig 3⇓). A high signal on T2 weighted sequences is seen in all diseases with alveolar exudates, making MRI ideal for imaging pneumonia.65 MRI has one major advantage over chest radiograph—it is a cross sectional imaging technique without the limitations of two dimensional imaging.3765 It can also demonstrate complications of pneumonia such as abscess or effusion.65
Evidence for the accuracy of MRI and interobserver agreement in interpreting pneumonia in children is sparse. Only three studies have compared MRI for the detection of lung consolidation against computed tomography as the gold standard (table 2⇓).647071 Other studies have compared MRI findings with chest radiography; however, these studies provide weak evidence because chest radiography is considered an inadequate gold standard (table 2⇓).
Tele-reading of diagnostic images by expert radiologists
Many low and middle income countries have few or no radiologists, and even fewer pediatric radiologists, which hampers the reliable interpretation of chest images.72 However, tele-radiology may overcome the lack of radiology expertise through electronic transfer of digital imaging for reviewing remotely.337273
Tele-radiology consultation is feasible and may help obtain a more accurate radiological diagnosis.72 Several free platforms are available to transfer chest radiographs, ultrasound scans, or even larger datasets to experts for interpretation.73 Challenges include generating high quality images at the source, the digital transfer chain, limited expertise of the reporting radiologist,73 language barriers, legal problems, and image transfer sustainability.72
Various diagnostic modalities are available to support the diagnosis of the cause of pneumonia. These include tests done on respiratory specimens, such as microscopy (including targeted fluorescence microscopy for specific pathogens), microbiological and viral culture, antigen detection, and molecular testing for nucleic acid from the pathogen. Some of these tests may be used with other specimen types—for example, molecular or antigen testing of urine samples. Blood culture remains important for identifying bacteremic pneumonia, and serological testing may be useful for specific pathogens or for retrospective confirmation of infection.
Improved methods for microbiological diagnosis have challenged traditional assumptions regarding the causes of pneumonia and the interpretation of microbiological results. Multiple potential pathogens are often detected in respiratory samples from children with pneumonia, and coinfection may be particularly important in severe disease. Furthermore, some organisms such as Mycobacterium tuberculosis, which were formerly regarded as causing chronic disease, have been shown to be associated with acute pneumonia in tuberculosis endemic areas.74
Limitations of current microbiological testing methods
Accurate identification of the cause of pneumonia is needed to guide appropriate treatment, to develop rational guidelines for empiric treatment, and to evaluate the impact of preventive interventions, such as vaccination. However, it can be difficult to attribute the cause of pneumonia because disease may be caused by multiple pathogens, bacteria seldom invade sterile sites (such as the bloodstream), and it is hard to distinguish colonizing flora from pathogenic organisms in respiratory samples. Although there have been important advances in rapid and accurate identification of potential pathogens using multiplex molecular methods or rapid antigen detection, the evaluation of novel diagnostic tests is hampered by the lack of an accurate reference standard test for assigning the cause of pneumonia or for distinguishing between bacterial and viral pneumonia.
Current guidelines for microbiological diagnosis
Given the difficulties associated with the interpretation of etiologic diagnostics and lack of data on their impact on management, BTS guidelines for community acquired pneumonia (CAP) recommend microbiological testing only in children with severe pneumonia who need to be admitted to intensive care or those with complications.28 In such cases, a variety of diagnostic tests are suggested including blood culture; nasal specimens for viral detection; acute and convalescent serology for viruses and atypical organisms; and pleural fluid (if present) for microscopy, culture, antigen detection, and polymerase chain reaction. These guidelines are similar to the South African Thoracic Society guidelines, which recommend etiologic diagnostic testing only in children who are admitted to hospital; tests include blood culture, pleural fluid testing, possible viral testing on nasal samples, and microbiologic testing for tuberculosis when clinically suspected using induced sputum or other specimens.31 However, these guidelines differ substantially from those of the Pediatric Infectious Diseases Society and the Infectious Diseases Society of America (IDSA), which include rapid diagnostic testing for influenza virus and other respiratory viruses in children with CAP who are being treated in the community or in hospital.30 For children with CAP who are in hospital, IDSA guidelines further recommend blood culture in those with complicated or moderate to severe disease and sputum for Gram stain and culture in those who can produce a specimen.30 Tracheal aspirates for culture and detection of viral pathogens are recommended in children who have been intubated, whereas bronchoscopy, bronchoalveolar lavage, or lung biopsy are recommended only in those with severe pneumonia who have negative diagnostic tests.30 In addition to testing children with severe disease, targeted testing may be warranted in outbreak situations, or if infection with specific pathogens, such as M tuberculosis or Bordetella pertussis, is suspected.
Value of different specimen types and associated diagnostic tests
Optimal specimen collection is key to accurate identification of the cause of pneumonia. Diagnostic yield depends on the type and quality of the specimen collected, timing of collection, previous use of antimicrobials, transport, storage, and laboratory processes for testing. It may be difficult to obtain a representative sample from the lower respiratory tract, and contamination with organisms such as Streptococcus pneumoniae or Haemophilus influenzae, which colonize the upper airways but are also important pathogens, is likely. This may lead to overestimation or underestimation of the contribution of these bacteria to pneumonia. Similarly, although molecular diagnostics enable the detection of many respiratory viruses and bacteria that are not easy to culture, it is difficult to distinguish live pathogenic organisms from those that are colonizing or infecting the upper respiratory tract or nucleic acids left over from previous infections.
A range of specimens may be useful for investigating the causes of pneumonia, and the diagnostic utility varies with specimen type and specific organism (table 3⇓).7576 Nasopharyngeal samples, swabs or aspirates, have traditionally been used for the detection of most viruses and for pertussis. Flocked nasopharyngeal swabs provide a similar yield to aspirates, are simpler to obtain than aspirates,7778 and may provide a better yield than other swab types.798081 Studies are inconsistent regarding the relative yield of nasopharyngeal versus oropharyngeal specimens for the detection of viruses; there may be incremental yield from testing of both specimen types for specific viruses such as influenza A or adenovirus.82838485 Nasopharyngeal specimens may also be useful for the diagnosis of tuberculosis using the polymerase chain reaction (PCR) in children from whom it is not possible to obtain more representative lower respiratory tract samples.86
Although sputum induction has been shown in many studies to be safe and to provide an effective specimen for microbiologic diagnosis of pulmonary tuberculosis in children of all ages, there are relatively few studies of its use in acute pneumonia. Recent data suggest that the diagnostic yield of induced sputum may be greater than for nasopharyngeal samples for specific organisms. In a multicenter case-control study of children in hospital for severe or very severe pneumonia, the use of induced sputum samples increased the detection of B pertussis using PCR by 48% compared with the use of a combined nasopharyngeal and oropharyngeal swab.87 Two South African studies also found an increased yield of B pertussis on PCR with induced sputum,1188 although no difference was shown in another hospital based study.89 Several studies have shown good agreement between PCR testing of induced sputum and nasopharyngeal swabs for viruses; the yield is highest when both sample types are tested.11757990 The use of induced sputum has also been reported to be helpful for microbiologic confirmation of Pneumocystis jirovecii pneumonia in HIV infected children when using PCR for diagnosis.91 However, as with nasopharyngeal samples, induced sputum samples are likely to be contaminated with upper respiratory tract flora, so it may be difficult to interpret results.
Nasopharyngeal and induced sputum samples may be cultured for bacteria and fungi, although culture is increasingly being replaced (or supplemented) with nucleic acid amplification tests for a range of microbes, or by rapid viral antigen detection. If samples are tested for a wide variety of micro-organisms using multiplex assays, a potential pathogen may be identified in almost all children with pneumonia. For example, a Finnish study reported that a potential bacterial or viral pathogen was detected by PCR in 97% of children in hospital with pneumonia using an adequate induced sputum sample,12 and in a South African study, both bacteria and viruses were detected in nasopharyngeal samples from 87% of children with pneumonia.11
However, this high yield of apparent pathogens should be interpreted with caution. Most studies reporting on the use of molecular testing of respiratory tract samples have been case series without relevant matched controls. A large meta-analysis of case-control studies of viral detection in children with LRTI reported that respiratory syncytial virus (RSV) was most strongly associated with LRTI (odds ratio 9.8), whereas influenza virus (5.1), parainfluenza virus (3.4), and human metapneumovirus (3.8) were less strongly associated. Rhinovirus was only weakly associated with LRTI (1.4), whereas no association was found for adenovirus, bocavirus, or coronaviruses.92
A Dutch study found that Mycoplasma pneumoniae DNA was not detected any more often in children with any respiratory tract infection than in control children.93 Two recent studies, a birth cohort in South Africa11 and a multicenter study in the US,94 have reported similar findings. RSV, influenza virus, and B pertussis have been consistently and strongly associated with pneumonia in many studies. However, there is considerable heterogeneity with regard to the causative role of many other viruses and bacteria detected in respiratory samples, which may relate to case definitions, sampling, and the inclusion of a relevant control group.92
Few studies have evaluated the impact of rapid viral diagnostics on patient outcomes or healthcare utilization. A randomized controlled trial (RCT) of 418 patients (2-21 years of age) presenting to an emergency department with influenza-like symptoms randomized patients to groups where physicians did or did not receive results of rapid influenza viral testing.95 The number of ancillary investigations (P<0.001 for complete blood count and blood culture; P=0.011 for urine analysis and urine culture) and chest radiographs performed (P=0.001) were significantly reduced in patients who tested positive. These patients were also prescribed fewer antibiotics (P<0.001) and had a shorter length of stay (P<0.001). A second study randomized 204 children (3-36 months of age) presenting with febrile acute respiratory illness at an emergency department to rapid viral testing or routine care (which included rapid viral testing if specifically requested by the physician).96 No difference was seen in length of visit, rate of ancillary testing, or prescription, although the number of post discharge antibiotic prescriptions was reduced in the group who had viral testing (relative risk 0.36; 95% confidence interval 0.14 to 0.95). Two retrospective record review studies showed a reduction in antibiotic usage in children who received rapid viral testing.9798 No large studies have evaluated the impact of rapid viral testing and the associated reduction in antibiotic use on patient outcomes. Given that bacteria and viruses are often detected at the same time, and that children admitted to intensive care with severe disease and RSV infection may have concomitant bacterial infection,99 rapid viral testing cannot be used alone to rule out bacterial infection, particularly in children with severe pneumonia.
Overall, these results call into question the routine use and interpretation of costly, multiplex detection assays and suggest that more limited assays that target specific pathogens that are strongly associated with pneumonia, such as influenza, RSV, and pertussis, may be more appropriate (table 4⇓).
Real time PCR assays can often estimate the density of colonization (pathogen load) in the upper respiratory tract, which may be associated with the likelihood of infection. For example, the pneumococcal colonization density has been associated with respiratory virus infection and invasive pneumococcal pneumonia.100 However, pathogen density in the upper respiratory tract was not associated with pneumonia in a recent case-control study.11
Reducing contamination of samples
Samples that are less likely to be contaminated by upper respiratory flora may be less prone to problems of interpretation. Bronchoscopy with bronchoalveolar lavage (BAL) or aspirate may provide a representative sample from the lower respiratory tract but it is invasive and requires specialized skill and sedation of the child. Few studies have compared the viral yield from BAL with that from upper respiratory samples, but the results seem to be similar.101 BAL is mainly used in patients with hospital acquired infections, immunocompromised patients, and those with severe or refractory disease when other specimens have not been able to identify the causative agent.
Transthoracic lung aspiration provides a representative specimen, uncontaminated by the upper airway flora, when pneumonia is peripheral and accessible to needle aspiration; however, it is invasive and requires expertise. Recent studies and a review have provided reassuring data on the safety of this procedure.76 In a report from the Gambia, the diagnostic yield depended on the method of testing, with a potential pathogen identified by a combination of culture and molecular testing in 53 of 56 samples (including 47 lung aspirates and nine pleural fluids).102Streptococcus pneumoniae was identified in 91% of samples by molecular testing, but only 25% of samples by culture.102 Because case-control studies are not possible with this specimen type, inferences regarding causality are not straightforward. It seems probable that contamination with upper respiratory tract flora is less likely with lung aspirates; however, highly sensitive nucleic acid amplification assays may detect very small amounts of nucleic acid and do not necessarily imply the presence of viable organisms in large numbers.
Because the yield from blood culture is typically low, both BTS and IDSA guidelines recommend against routine collection of blood cultures in ambulatory children; blood culture is indicated only in children who are in hospital,30 in intensive care,28 or have complicated pneumonia.2830 Most studies have not evaluated the yield of systematically collected blood cultures from all children presenting with pneumonia but have been retrospective studies of existing practice, which may be biased by collection in sicker children.
Two recent multicenter studies of children in hospital with pneumonia in the US that were performed after implementation of PCV, documented notably different rates of bacteremia—1.5% (6/390)103 and 7% (26/369)104—with S pneumoniae the predominant pathogen in both. Collection of a blood culture was associated with increased time to discharge, even after matching on propensity scores based on clinical factors.103105 Cost effectiveness modeling predicts that, in the US, blood cultures would need to be drawn from 118 children in hospital to identify one child with bacteremia in whom the result would lead to a meaningful change in antibiotic treatment.106 It may be more cost effective to target testing to children at higher risk of bacteremia, such as those less than 6 months of age with fever; those admitted to intensive care; immunosuppressed patients; those with a central line in place; or those with chronic medical conditions, effusion, or empyema.106
Several studies have described substantially enhanced detection of S pneumoniae in blood samples using real time PCR targeting the lytA gene of the pneumococcus, compared with blood culture.107108109 However recent results from the multicenter Pneumonia Etiology Research for Child Health (PERCH) study showed both poor sensitivity (64%; 36/56 microbiologically confirmed pneumococcal pneumonia cases were PCR positive) and specificity (5.5%; 273/4987 of non-pneumonia controls were PCR positive).110 These results illustrate the need for well controlled studies of the cause of pneumonia, and the major limitations of studies of case series in correctly ascribing the cause.
Serological testing of blood specimens may be useful for epidemiological studies of the cause of pneumonia, where comparison of acute and convalescent titers may help identify the causative agent,93111112 although the diagnostic value of a single specimen collected in the acute phase of the illness is generally limited. Pertussis serology may be a useful complementary diagnostic tool in older children and during the later phases of illness.113114
Because urine is relatively easy to obtain from most children, pathogen specific nucleic acids or antigens in urine are attractive diagnostic targets. Several prospective studies and a systematic review have concluded that a simple immunochromatographic lateral flow test for pneumococcal C polysaccharide antigens in urine has high sensitivity (close to 100%) for invasive pneumococcal infection but lacks specificity (60-80%), particularly in children with nasopharyngeal pneumococcal carriage.115116117 More recently, an immunodiagnostic assay for serotype specific detection of polysaccharides from 13 different pneumococcal serotypes (those found in PCV13) has been developed,118 and it has been shown to be sensitive and specific (98% and 100%, respectively, when using bacteremic episodes with one of the 13 serotypes included in the test as the reference) for pneumococcal pneumonia in adults.119 However, this test is likely to also be positive in people with nasopharyngeal carriage of pneumococci. Urine lipoarabinomannan testing by lateral flow assay has been shown to have clinical utility in adults with tuberculosis and advanced HIV infection, but it lacks both sensitivity and specificity in children with tuberculosis.120
Host biomarkers for distinguishing between bacterial and viral infection
In the absence of tests that can identify the specific cause of pneumonia, a test that could discriminate between children with bacterial pneumonia, who require antibiotics, and those with viral pneumonia, would be a major advance. Host biomarkers have shown some promise in this regard; however, their evaluation is complex given the lack of an accurate reference standard comparator test.
The most commonly used biomarkers, procalcitonin (PCT) and C reactive protein (CRP) seem to have suboptimal sensitivity and specificity for identifying children with bacterial pneumonia, with no satisfactory cut-off point on receiver operating characteristic curve analysis.121122123 The reported performance characteristics of procalcitonin versus CRP vary, with no clear evidence that either is superior.121124
BTS guidelines recommend against the use of acute phase reactants for distinguishing between bacterial and viral pneumonia,28 whereas IDSA guidelines recommend that they are not used as the only distinguishing determinant.30 However, two European RCTs showed a reduction in antibiotic use in children with lower respiratory tract infection randomized to a procalcitonin guided algorithm versus standard of care. One study that randomized 319 pediatric inpatients showed a reduction in the prescription of antibiotics (85.8% v 100%; P<0.05) in the procalcitonin group compared with the control group.125 The other study, which randomized 337 children attending an emergency department, failed to show differences in prescribing rates but showed that the mean duration of antibiotic exposure was reduced in the procalcitonin group (−1.8 days, 1 to −0.5).126 A recent Vietnamese RCT compared POC testing for CRP with routine care in of 2037 patients (including 1028 children) with non-severe acute respiratory tract infection. It found a reduction in immediate antibiotic use in children (and adults) who received CRP testing (odds ratio 0.49, 40 to 0.61) with no increase in adverse events.127
Semi-quantitative POC testing for CRP or procalcitonin is feasible and available but requires an instrument to interpret results and, in the case of procalcitonin, remains relatively costly (even though it may be cost effective in critically ill adult patients).128 The safety of such approaches needs to be established in low and middle income countries with higher rates of complicated and severe pneumonia, its cost effectiveness assessed, and appropriate procalcitonin cut-off values for pediatric pneumonia identified.
White blood cell indicators are less reliable than procalcitonin or CRP for identifying children with bacterial pneumonia.122124129 Several other biomarkers, including the combination of haptoglobin, tumor necrosis factor (TNF) receptor 2, or interleukin 10, and tissue inhibitor of metalloproteinases 1,130 have shown promise for the classification of children with bacterial pneumonia but require further evaluation.
Current research on the respiratory microbiota in health and during infection is likely to refine our understanding of how ecological changes in this niche may predispose to or predict pneumonia. Tests that can identify such changes may allow us to identify children at risk of pneumonia and help discriminate between pneumonia caused by different classes of pathogens.
Work is needed to clarify the role of infection with multiple pathogens in the pathogenesis of pneumonia, to inform the interpretation of multiplex diagnostic assays. However, reliable evaluation of the accuracy of diagnostics relies on the availability of accurate reference standard tests for the cause of pneumonia. In the absence of such reference standards, the development of consensus guidelines for the interpretation of microbiological testing would be useful, both to guide clinical decision making and for studies of the causes of pneumonia.
An increasing area of interest is the use of blood specimens to identify a host transcriptional signature associated with specific pathogens. A recent analysis of transcriptional data from multiple cohorts identified a transcriptional signature that distinguished influenza virus infection from bacterial and other viral infections.131
Similarly, a whole blood transcriptional profile that distinguishes RSV infection from other viral infections and predicts the severity of disease has been described.132 A two transcript signature has recently been identified which holds promise for distinguishing between febrile children with bacterial infection and those with viral infection.133 Although proof of principle has therefore been established for transcriptional profiles, there are considerable challenges in translating these findings to the diagnosis of the cause of pneumonia. Signatures require validation in different geographic settings with different co-prevalent illnesses and human populations, and the considerable technical obstacle of gene expression analysis on multiple markers in an inexpensive and portable format must be overcome.
Greater use of POC ultrasound in primary care settings may offer useful radiological imaging for pneumonia, particularly when combined with tele-radiology. Further study of rapid sequence chest MRI, especially for the detection of complications, is warranted. Pneumonia occurs because of dysbiosis in the normal microbial ecology of the respiratory tract in a susceptible child. Improvements in specimen collection and microbiological testing have enabled the detection of many potential organisms from respiratory specimens, but the understanding of how these interact in the pathogenesis of pneumonia is limited. Many new microbiological tests have high sensitivity for detecting potential pathogens but are limited in their ability to attribute causation, particularly when testing upper respiratory tract specimens. Targeted testing approaches for specific pathogens that are strongly associated with pneumonia may offer a more useful strategy than highly multiplexed testing, particularly in specific patient groups. A rapid, affordable, and reliable test to distinguish bacterial pathogens in pneumonia is urgently needed.
Questions for future research
What place does chest ultrasound have in the screening and diagnosis of childhood pneumonia?
What ultrasound findings are most accurate for the diagnosis of pneumonia?
Is ultrasound useful for informing antibiotic treatment for pneumonia or monitoring treatment response?
What is the role of magnetic resonance imaging (MRI) for the diagnosis of pneumonia?
Can MRI accurately distinguish between lymphadenopathy caused by tuberculosis and that caused by bacterial or viral lower respiratory tract infection?
What is the optimal sampling strategy for the detection of pathogens in childhood pneumonia?
What biomarker or biomarker signature accurately distinguishes bacterial or mixed viral-bacterial pneumonia from viral pneumonia?
What is the impact of rapid point-of-care biomarker testing on antimicrobial prescription rates and clinical outcome in children with severe pneumonia and children with pneumonia in lower and middle income countries?
What role does infection with multiple pathogens play in the development and severity of pneumonia?
What is the impact of rapid multiplex testing for respiratory pathogens on patient and health system relevant outcomes?
Series explanation: State of the Art Reviews are commissioned on the basis of their relevance to academics and specialists in the US and internationally. For this reason they are written predominantly by US authors
Contributors: All three authors helped in the conception, design, drafting, final approval, and accountability of this article’s accuracy and integrity. HJZ serves as the overall guarantor.
Competing interests: All authors have read and understood BMJ policy on declaration of interests and declare the following interests: None.
Provenance and peer review: Commissioned; externally peer reviewed.
Patient involvement: No patients were involved in the creation of this article.