Globally, pneumococcal infections, caused by Streptococcus pneumoniae (Pneumococcus), are one of the leading causes of morbidity and mortality in children <5 years of age ¹ . A variety of clinical syndromes of varying severity are associated with pneumococcus, including pneumonia, meningitis, bacteremia, otitis media and sinusitis ² . It has been estimated that prior to the introduction of pneumococcal conjugate vaccines (PCVs), diseases caused by pneumococcus were responsible for approximately 600,000 deaths per year globally in children 1-59 months of age ³ .

Pneumonia is among the leading causes of mortality in children under 5 years of age ^{1, 4} . The main causative pathogens attributable to pneumonia include Streptococcus pneumoniae, Haemophilus influenzae type B, all of which have vaccine-preventable bacterial causes, and respiratory syncytial virus ⁵ . Infants and young children are at highest risk for serious disease ⁶ , with children younger than 4 months being more likely to die ⁷ . In addition to pneumococcus, a variety of other infectious agents are related to pneumonia in children.

In the last two decades, more than 140 countries globally have introduced PCVs into national routine immunization schedules. Several studies have demonstrated the impact of PCVs on reducing invasive pneumococcal diseases and hospitalizations due to pneumonias ^{2, 8, 9} . However, pneumonia mortality is the greatest concern for policymakers and donors, and there is limited evidence on the impact of PCVs on pneumonia deaths in children.

Recent evidence from countries in Latin American using secondary mortality data demonstrated the impact of PCVs on pneumonia mortality in children under 5 years of age ^{10– 13} . Most studies did not include children <3 months of age assuming that perinatal causes of mortality and other etiologic agents and not pneumococcal disease are responsible for the pneumonia mortality in this age group. Nonetheless, this assumption is not fully backed up by the very little available evidence in the literature on the etiology of pneumonia in this age group. Although selected studies have indicated that respiratory viruses are the most common pathogens of pneumonia in infants and toddlers, some investigators have implicated pneumococcus and Haemophilus in 4–20% of cases. These findings vary significantly in developing versus industrialized countries, over time, and depending on laboratory methods used to assess etiologies.

It is still not clear whether pneumococcus is a significant cause of pneumonia in younger children, particularly neonates and children <3 months of age. Whether or not to include children in this age group in impact assessment studies will depend on evidence suggesting whether pneumococcus is a significant etiology of pneumonia and thus an important burden in children under 3 months of age.

This systematic review aims at summarizing the evidence of the bacterial etiology of respiratory infections in children under 3 months of age, in particular the role of pneumococcus as a significant etiology in this age group.

The study protocol was published in PROSPERO under registration number CRD42020158091. We followed PRISMA recommendations ¹⁴ , and a completed PRISMA checklist is provided as extended data ¹⁵ .

A total of 4,313 studies were retrieved in searches. After eliminating duplicates, a total of 3,744 references were screened by title and abstract review. Out of the 602 selected citations, a further 580 were excluded in two rounds of review, with 22 remaining papers eligible for data extraction reporting on 13 studies included in this review ( Figure 1). A complete list of reasons for excluding studies as well as references are provided in the extended data. Out of correspondences to authors of 54 studies, 16 responses were received, but new data was obtained for only 9 studies, which are included in 22 selected papers. Detailed list of included papers by database can be found in the extended data ¹⁵ .

13 (n=13) studies were considered in this review ( Table 1). Studies range over four decades (1980-2020), present different study designs, and consider a wide variation of number of children enrolled and assessed. Despite target age group being younger than 3 months of age, two studies evaluated only neonates aged ≤28 days, while some studies evaluated children <3 months including neonates and others excluded neonates from the study, thus including only children >28 days to <3 months.

While one study only assessed viral etiologies ¹⁶ , three studies evaluated only bacterial etiologies ^{17– 20} . Most studies report on blood (n=12) and nasopharyngeal swabs/aspirates or nasal washing (n=8), although some studies also collected other specimens, including cerebrospinal fluid (CSF), urine, broncho-alveolar lavage (BAL), pleural effusion and lung biopsies. Added to this variability of samples available and tests conducted, there were varying methods used and for etiologic diagnosis, with most studies reporting cultures, but some also using serology and antigen testing for viral infections. Most recent studies ^{16, 21– 33} included molecular techniques, with known increased sensitivity (i.e., ability to detect pathogens) to identify many etiologic agents. The variety of study designs and methods used in the studies made it inappropriate to conduct a meta-analysis.

Etiologies identified by each study have, as expected, varied greatly considering study characteristics and methods as described above. In Table 2 below, main study characteristics and etiologic agents identified are presented ( Table 2).

The only study conducted in the 1980s ³⁴ in India, evaluated only neonates up to 28 days. This study of 44 neonates reports no Streptococcus agalactiae as an aetiologic agent of neonatal pneumonia, but rather demonstrates a high proportion of S. pneumoniae (22.4%) (using antigen testing), and Gram-negative agents (25%).

A multinational World Health Organization (WHO) study in three countries to assess etiology of severe disease in children < 3 months of age, considering also but not restricted to pneumonia, was conducted from 1991 to 1993 ³⁵ . This study is reported in different papers, one of which describes results in all countries combined ³⁵ , and country specific results, namely Ethiopia ³⁶ and Papua New Guinea ^{37, 38} , as there are some variations in methods in each study site. This was the largest prospective study of early infant infections in developing countries, and it also reported on the absence of Streptococcus agalactiae and the importance of Gram-positive (61%) and Gram-negative (24%) organisms among the 167 positive blood cultures with isolates identified. Among these, S. pneumoniae, S. aureus and S. pyogenes (Gram-positives), and E. coli and Salmonella (Gram-negatives) are the most noteworthy. Viral etiologies were also important pneumonia agents, as reported in Ethiopia ³⁶ .

A retrospective laboratory-based surveillance study including children with bacterial invasive disease was conducted in Korea, and two papers report on different study periods, from 1996-2005 ¹⁸ , and from 2006-2010 ¹⁹ . This study only considered bacterial etiologies from blood, PPE, CSF, and here S. agalactiae is frequently isolated in both study periods.

Another study evaluating only neonates up to 28 days of age in China from 2006-2008 ²⁰ also demonstrated a significant proportion of Gram-negative agents in children hospitalized with community acquired pneumonia, mostly K. pneumonia and E. Coli.

One study evaluated viral etiologies of hospitalized children with respiratory infections only ¹⁶ . Children aged up to 28 days were excluded, so only children aged over 28 days and under 3 months of age were considered. Molecular methods were used to assess etiologies in nasopharyngeal aspirates (NPA) (no other specimen was obtained and evaluated). A high proportion of respiratory syncytial virus (RSV) was observed in children with positive isolates.

Nascimento-Carvalho et al. ^{26– 29} in multiple prospective cross-sectional studies conducted in Brazil during 2003–2005 and evaluating 19 different etiologies in blood and NPA using culture, serology and PCR, report the importance of viral etiologies in children hospitalized with community acquired pneumonia. Here, children aged 28 days and younger were excluded from the study, so data presented is for children from 1 to 3 months of age.

Finally, a large WHO multinational study in seven countries to assess the etiology of severe pneumonia, using a case-control design and including hundreds of children, was conducted from 2011-2014 ^{21, 24, 25, 31– 33} . Children aged 28 days and younger were also excluded, and various specimen types and multiple laboratory methods were used to identify viral and bacterial etiologies of severe pneumonia. Findings also reinforce the importance of viral etiologies in children aged 1–3 months, but bacterial agents, particularly S. pneumoniae and S. aureus, were also reported as relevant agents.

More recent studies have similar methods, including prospective cohort designs, the use of molecular methods, and collection of various specimens including blood and NPA at a minimum ^{22, 23, 30} . While the studies in Bhutan ²³ and Bolivia ²² reported a significant proportion of viral etiologies, particularly RSV and rhinovirus, the study in Malaysia ³⁰ reported a high proportion of Gram-positive bacterial agents, particularly S. aureus, H. influenzae, and S. pneumoniae, isolated or combined with viral etiologies.

The risk of bias assessment of the studies is presented in Table 3. In general, studies presented high risk of bias, particularly due to design, small number of subjects, specimens collected, and laboratory methods. Many of them were conducted in different decades, when availability and accuracy of diagnostic tools varied significantly. Very few studies included controls, namely the multinational WHO Young Infant Study Group (1999) ³⁵ , and PERCH Study ^{21, 24, 25, 31– 33} . Given the paucity of evidence, we opted to report on all studies, and consider their limitations and potential biases in interpreting the results.

Pneumonia is a very frequent childhood disease leading to disease burden significantly higher in children when compared to other age groups. Several studies, many of which conducted in developing countries, have evaluated pneumonia etiologies in the past decades. Understanding pneumonia etiology is key to guiding diagnosis and management approaches to pediatric pneumonia.

It is well known that pneumonia etiology varies by age. Nonetheless, very limited evidence is available for young children, particularly children under 3 months of age. Determining the etiology of community acquired pneumonia in children, including severe disease leading to hospitalizations, is very important not only to define treatment guidelines, but also to implement preventive strategies at national level, and more recently to also assess the impact of selected interventions, as pneumococcal conjugate vaccines (PCVs) are introduced.

Identifying the cause of pneumonia in children is difficult because of varying syndromic presentations, challenges in obtaining specimens for laboratory assessment, and the lack of rapid, commercially available, accurate laboratory tests for most pathogens, among others. A recent landscape assessment and literature review conducted by Gilani et al. ³⁹ reported that published or ongoing (at the time) studies of pneumonia etiology in children present a multiplicity of case definitions, levels of clinician involvement, facility types, specimen collection, and laboratory techniques, thus reinforcing the need for the standardization of methods and analyses of pneumonia etiology in children.

Limited reviews have reported on the etiology of pneumonia in children, mostly in specific locations and in younger than 5 years of age ⁴⁰ . Studies conducted in developed countries clearly demonstrate that the pattern of etiologic agents causing pneumonia in children, in particular severe pneumonia, has been changing over the past decades. While in the 1980s bacterial agents including Staphylococcus bacteria ( aureus and pyogenes) were the main causative agents of severe pneumonia in children, over time studies began reporting an increase in the proportion of Gram-negative agents and Group B Streptococcus ( S. agalactiae), which accounted for most pneumonia cases in children. Also, there is growing evidence demonstrating the importance of viral etiologies, including RSV, rhinovirus, influenza, parainfluenza, alone or in combination with bacterial pathogens, as important etiologies of pneumonia in children.

In the Canadian Guidelines for treatment of pediatric pneumonia from 1997 ¹⁷ , the reported main pathogens causing pneumonia in infants aged 1-3 months are, in order of frequency: Chlamydia trachomatis, RSV, other respiratory viruses, and Bordetella pertussis.

In a review article published in 2002, McIntosh ⁵ reports on the bacterial and viral agents causing pneumonia in children, particularly S. pneumoniae and S. pyogenes, S. aureus and H. influenza among bacterial agents, and RSV, influenza, parainfluenza, adenovirus, and rhinovirus among viral agents. McIntosh ⁵ reinforces that for treatment decision making, one should first consider the age of the child. To that end, no comprehensive review has been conducted on the etiology of pneumonia in children under 3 months of age.

This systematic review included 13 studies reported in 22 publications, conducted from 1986 through 2020 in a variety of locations, mainly in developing countries. Results were variable, depending on time in which study was conducted, study design, and laboratory methods used.

Earlier studies conducted in the 80s and 90s ^{34– 37} demonstrate that S. pneumoniae is a very important etiologic agent even in neonates. Viral etiologies including RSV, influenza and parainfluenza were also observed in the WHO multicenter prospective study ^{35– 37} .

A retrospective study in Korea ^{18, 19} conducted over 1996 to 2010 reported S. agalactiae as a significant agent. Nonetheless, these studies were severely biased for various reasons. First, the study was retrospective and based on laboratory surveillance, with no clinical information of patients enrolled, but rather considering invasive disease as of pulmonary focus when pulmonary or pleural specimens had been obtained. In addition, this study included both community and nosocomial infections, it not being possible to disaggregate them. Finally, only bacteria were evaluated and no viral etiologies. This study also included children younger than 28 days.

Another retrospective cohort study conducted in China ²⁰ evaluated only bacterial etiology on young neonates aged ≤28 days. Furthermore, only sputum specimens were collected and processed, which imposes major biases in this study as well.

More recent research conducted in the past 15 years, using prospective cohort designs and better standardized methods and case definitions, and applying molecular diagnostic techniques to evaluate etiologic agents including bacterial and viral etiologies, suggests a significant proportion of viral agents causing pneumonia in children younger than 3 months of age ^{16, 21– 32, 33} . Of note, Finianos et al. ¹⁶ in Lebanon evaluated viral agents only. Nathan et al. ³⁰ in Malaysia also report a significant proportion of bacterial etiologies, particularly S. pneumonia, H. influenza and S. aureus. This was also reported by the most robust body of evidence to date on the etiology of hospitalized pneumonia in children, resulting from a WHO multinational cohort study conducted in 7 study sites from 2011-2014, the Pneumonia Etiology Research for Child Health (PERCH) study ^{21, 24, 25, 31– 33} .

This review demonstrates that available evidence on etiology of pneumonia in young children, particularly children younger than 3 months of age is based on a variety of studies with non-standardized methodology. Syndromes and case definitions as well as age subgroups included (younger than 7 days and younger than 28 days) vary significantly among studies. Samples collected and tests performed also vary significantly, and also over time, with molecular methods available in more recent studies. Studies also vary in terms of sample size, and time and locality in which it has been conducted. All of these are known factors which may influence the reported etiology and also the ability to identify selected agents. Adequate specimens and testing methods should be used for studies evaluating etiology of pneumonia in children, in particular molecular techniques with higher sensitivity.

Despite the above limitations and challenges, this review reinforces that Gram-positive organisms, in particular S. pneumoniae, are still important etiologic agents of pneumonia in children under 3 months of age and should thus be considered when assessing impact of PCV in the children. In addition, viral etiologies are also important, responding for a significant proportion of pneumonia in children younger than 3 months of age.