Antibiotics for Paediatric Community-Acquired Pneumonia: What is the Optimal Course Duration?

Abstract

Despite significant global reductions in cases of pneumonia during the last 3 decades, pneumonia remains the leading cause of post-neonatal mortality in children aged <5 years. Beyond the immediate disease burden it imposes, pneumonia contributes to long-term morbidity, including lung function deficits and bronchiectasis. Viruses are the most common cause of childhood pneumonia, but bacteria also play a crucial role. However, the optimal duration of antibiotic therapy for bacterial pneumonia remains uncertain in both low- and middle-income countries and in high-income countries. Knowing the optimal duration of antibiotic therapy for pneumonia is crucial for effective antimicrobial stewardship. This is especially important as concerns mount over rising antibiotic resistance in respiratory bacterial pathogens, which increases the risk of treatment failure. Numerous studies have focused on the duration of oral antibiotics and short-term outcomes, such as clinical cure and mortality. In contrast, only one study has examined both intravenous and oral antibiotics and their impact on long-term respiratory outcomes following pneumonia hospitalisation. However, study findings may be influenced by their inclusion criteria when children unlikely to have bacterial pneumonia are included. Efforts to differentiate between bacterial and non-bacterial pneumonia continue, but a validated, accurate, and simple point-of-care diagnostic test remains elusive. Without certainty that a child has bacterial pneumonia, determining the optimal duration of antibiotic treatment is challenging. This review examines the evidence for the recommended duration of antibiotics for treating uncomplicated pneumonia in otherwise healthy children and concludes that the question of duration is unresolved.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40272-024-00680-4.

Key Points

Studies from resource-limited low- and middle-income countries support 3 days of oral amoxicillin for treating children aged 2−59 months with fast breathing and/or chest-indrawing pneumonia. However, questions remain over the reliability of the data, as many participants likely had non-bacterial pneumonia, potentially biasing results towards the null.

In high-income countries, 3−5 days of oral antibiotic regimen is noninferior to a 7−10-day course, particularly for achieving clinical cure for those treated as outpatients, but 3-day courses should be avoided for those hospitalised with severe community-acquired pneumonia.

Among children hospitalised with community-acquired pneumonia from high-risk populations, a 5−6-day antibiotic course is the treatment of choice for those who have shown early improvement after 1−3 days of parenteral therapy and who can then transition to oral antibiotics to complete their course of antibiotics.

Nevertheless, the optimal duration of antibiotic treatment of uncomplicated pneumonia remains unresolved, and recommendations on duration should be regarded as provisional because of the methodological limitations in the clinical trials from which they are drawn.

Introduction

Community-acquired pneumonia (CAP) is the leading global cause of death in children aged 1−59 months [1]. Despite substantial reductions in pneumonia deaths per year over the last 3 decades (1,940,000 in 1990; 900,000 in 2015; 693,000 in 2019; 502,000 in 2021) [1, 2], the World Health Organization's (WHO) target of fewer than three deaths per 1000 live births in this age group by 2025 is unlikely to be achieved [3]. In 2021, 99.8% of pneumonia deaths in children aged <5 years occurred in low- and middle-income countries (LMICs), and only 0.2% were in high-income countries (HICs) [2]. Although mortality is low in HICs, CAP accounts for a large proportion of healthcare visits and hospitalisations [4]. The estimated incidence of pneumonia and hospitalisation in HICs is 34−40 and 2−23 per 1000 children, respectively [5]. In the USA, CAP resulted in 32−50 outpatient visits per 1000 children aged 1−5 years annually [6]. The burden of hospitalised pneumonia is also greater in Indigenous children living in socially disadvantaged communities in HICs [5, 7, 8].

In addition, childhood pneumonia is associated with long-term respiratory sequelae, including impaired lung function in children and adults. Among 2438 Tasmanian children in Australia with lung function data at both 7 and 53 years of age, respectively, childhood pneumonia increased the risk of ‘early below average’ and ‘accelerated decline’ of their lung function trajectories (odds ratio [OR] 2.0; 95% confidence interval [CI] 1.2, 3.5) [9]. Early childhood pneumonia is also associated with lower forced expiratory volume in 1-second (FEV₁) and forced vital capacity (FVC) z-scores in Indigenous Australian children. These reductions were observed when pneumonia occurred before 3 years of age (FEV₁ β −0.42; 95% CI −0.79, −0.04; FVC β −0.62; 95% CI −1.14, −0.09) and within 3−5 years of the episode (FEV₁ β −0.50; 95% CI −0.88, −0.12; FVC β −0.63; 95% CI −1.17, −0.10) compared with children who never had pneumonia [10]. A systematic review of 14 studies involving 23,276 participants examined the relationship between early-life acute lower respiratory infections (ALRIs), including pneumonia, and lifelong lung function [11]. Nearly all studies identified lung function deficits in individuals with a history of early childhood ALRIs.

Moreover, children may also develop chronic suppurative lung disease or bronchiectasis as a complication of hospitalised pneumonia. A case–control study conducted in Central Australia found that Indigenous children hospitalised with pneumonia had a significantly higher likelihood of developing bronchiectasis than did age- and sex-matched Indigenous children hospitalised for other reasons (OR 15.2; 95% CI 4.4, 52.7) [12]. Notably, among 180 Indigenous children from Australia, Alaska, and New Zealand with either chronic suppurative lung disease or bronchiectasis, 95% had prior ALRI hospitalisations, with recurrent pneumonia being linked to developing bronchiectasis [13]. A national birth cohort study in England, Scotland, and Wales reported that, among 3589 participants, 913 (25%) with an ALRI during early childhood had an increased risk of premature adult death from respiratory disease by 73 years of age, compared with those without an ALRI during childhood (adjusted hazard ratio [HR] 1.93; 95% CI 1.10, 3.37; P = 0.021) [14]. Those hospitalised for an ALRI during childhood had the highest risk [14].

Despite the substantial disease burden associated with childhood CAP, uncertainties remain over the ideal treatment of bacterial pneumonia [15], including the optimal duration of antibiotic therapy [16–18]. The advantages of shorter antibiotic courses over longer ones include a lower risk of developing antibiotic resistance, better adherence, fewer adverse events, and lower costs [19]. Nevertheless, shorter courses may also increase the risk of treatment failure and, in young children, risk further damage to developing lungs, resulting in impaired lung function across the life course [11]. Meanwhile, there is growing global concern over antibiotic resistance, including in Streptococcus pneumoniae, the major respiratory bacterial pathogen in children and adults [20, 21]. Antibiotic consumption is an important driver of antibiotic resistance [22], and antimicrobial stewardship programmes seek to reduce antibiotic consumption and resistance by decreasing inappropriate prescribing [23]. This includes the right patient receiving the right antibiotic in the right dose by the right route for the right time, with duration determined by the shortest effective course. Our narrative review begins by describing the clinical features of CAP and recommended antibiotic choices for treating uncomplicated CAP, then discusses antibiotic duration based upon evidence from the literature (for further details, see the table in the electronic supplementary material).

We searched PubMed for clinical trials from database inception to 1 September 2024 using the search terms ‘community-acquired pneumonia’, ‘antibiotic’, ‘randomised controlled trials’, and ‘child.’ We included studies that evaluated antibiotic treatment in children with uncomplicated CAP, primarily focusing on treatment duration. Macrolides were excluded, as they are not recommended as first-line therapies for childhood CAP unless penicillin allergy or Mycoplasma pneumoniae are suspected [24–29].

What is Community-Acquired Pneumonia?

There is no gold standard case definition for CAP in children [30]. For this review, we focused on studies where CAP was broadly defined as symptoms and signs of pneumonia caused by an infection acquired outside the hospital in a previously healthy child [24, 25]. These symptoms may include fever, tachypnoea, breathlessness, cough, or chest pain [25]. Although some centres perform chest radiographs (CXRs) to confirm the diagnosis, they are not recommended in outpatient settings [24, 25]. Uncomplicated pneumonia was defined as pneumonia without parapneumonic effusion, empyema, necrotising pneumonia, lung abscess, or systemic complications (e.g. multiorgan failure and acute respiratory distress syndrome) [31].

Given the considerable burden of childhood CAP in LMICs, the WHO simplified the definition to help community health workers more easily identify children with suspected pneumonia [26]. Any child aged 2−59 months with cough/difficulty breathing and fast breathing with/without chest indrawing is diagnosed with pneumonia. Severe pneumonia is associated with danger signs (inability to drink, persistent vomiting, convulsions, lethargy/unconsciousness, stridor in a calm child, or severe malnutrition) [26]. However, although the WHO definition is highly sensitive, it lacks specificity, fails to distinguish between bacterial and non-bacterial causes, and misclassifies other illnesses (e.g. asthma and malaria) as pneumonia [32]. The clinical and research definitions of pneumonia remain controversial, stressing the lack of a suitable definition of CAP that addresses these limitations [33].

Which Antibiotics Should Be Used?

Treatment is guided by the reported prevalence of and awareness by the treating clinician of various pathogens at different ages (e.g. S. pneumoniae in all age and risk groups, M. pneumoniae in school-aged children and adolescents), local epidemiology of antibiotic resistance (e.g. methicillin-resistant Staphylococcus aureus strains), illness severity (e.g. S. aureus), immunisation status (e.g. Haemophilus influenzae type b [Hib] if unvaccinated) and comorbidities (e.g. non-Hib strains in preterm neonates, malnutrition, underlying cardio-pulmonary disorders, or immunodeficiencies) [18, 34, 35].

In 2021, epidemiological data identified S. pneumoniae as the leading cause of pneumonia and pneumonia-related deaths in children. For example, in those aged <5 years, it was estimated to cause 11.6 million pneumonia episodes and 139,000 pneumonia-related deaths, corresponding to 31% and 28% of all pneumonia episodes and deaths in this age group, respectively [2]. Given its prevalence, S. pneumoniae remains a crucial target for antibiotic therapy, and amoxicillin, based upon pharmacokinetic and pharmacodynamic considerations, is the most active orally administered β-lactam against this pathogen [18, 36]. Consequently, various national and international guidelines, including those from the WHO, suggest oral amoxicillin as first-line therapy for uncomplicated CAP in previously healthy children who have received their age-appropriate vaccinations and are being treated as outpatients [24–29].

What is the Optimal Duration of Antibiotic Treatment?

Randomised Controlled Trials of Short-Term Outcomes

Short-term outcomes were clinical cure, treatment failure, relapse, and mortality within 30 days after randomisation.

Lower- and Middle-Income Countries

All the studies listed in the following sections recruited children aged 2−59 months from outpatient settings with either WHO-defined non-severe, fast breathing, or chest indrawing CAP.

The Pakistan MASCOT (Multicentre Amoxicillin Short-Course Therapy) randomised controlled trial (RCT) randomised 2000 children to either 3 or 5 days of oral amoxicillin 15 mg/kg three times daily (TID) [37]. There was no difference in treatment failure between the 3-day (209/1000 [21%]) and 5-day (202/1000 [20%]) arms (difference 0.7%; 95% CI −2.2, 4.2) [38]. However, only 259 of 1847 (14%) children with available CXRs had radiographic-proven CAP, and 54% were infants, which suggests that acute bronchiolitis was more likely [39].

In India, a similar finding was reported. The ISCAP (INDIACLEN Short-Course Amoxicillin Pneumonia) RCT randomised 2188 children to either 3 or 5 days of oral amoxicillin TID (31−54 mg/kg/day) [40]. On day 5 after randomisation, there was no difference in clinical cure between the 3-day (980/1095 [89.5%]) and 5-day (983/1093 [89.9%]) arms (difference 0.4%; 95% CI −2.1, 3.0).

Another Indian RCT undertaken 2 years later randomised 2009 children to either 3 days of oral amoxicillin TID (31−54 mg/kg/day) or 5 days of oral trimethoprim–sulfamethoxazole twice daily (BID) (7−11 mg/kg/day trimethoprim component) [41]. There was no difference in clinical failure between the amoxicillin arm on day 4 (137/993 [13.8%]) and the trimethoprim–sulfamethoxazole arm on day 6 (97/1016 [9.5%]) (absolute difference 0.04; 95% CI −0.035, 0.120). However, this open-label cluster RCT may have compromised internal validity because of potential selection and attrition bias [42].

The largest trial was conducted in Pakistan. It was another open-label cluster RCT, where 15,749 children were randomly assigned to either 3 days of oral amoxicillin BID (50 mg/kg/day) or 5 days of oral trimethoprim–sulfamethoxazole BID (8 mg/kg/day trimethoprim component) [43]. After adjusting for clustering, the risk of treatment failure was lower by day 4 for those receiving amoxicillin (326/9153 [3.6%]) than for those receiving trimethoprim–sulfamethoxazole by day 6 (592/6509 [9.1%]) (risk difference [RD] −5.5%; 95% CI −7.4, −3.7). Incomplete antibiotic treatment was high (amoxicillin 900/9153 [9.8%]; trimethoprim–sulfamethoxazole 1418/6509 [21.8%]). Furthermore, a multivariable analysis found that non-adherence predicted treatment failure (RD 2.9%; 95% CI 1.6, 4.1).

In Malawi, 3000 children without human immunodeficiency virus (HIV) infection but with chest-indrawing CAP with or without fast breathing were randomised to either 3 or 5 days of oral amoxicillin BID with age-group-determined doses (2−11 months = 500 mg/day; 12−35 months = 1000 mg/day; and 36−59 months = 1500 mg/day) [44]. By day 6, the 3-day arm was noninferior to the 5-day arm in treatment failure (85/1442 [5.9%] vs 75/1456 [5.2%], respectively; adjusted difference 0.7 percentage points, 95% CI −0.9, 2.4). However, the relative noninferiority margin of a 50% higher incidence of treatment failure in the 3-day versus the 5-day treatment arms was large and could miss an important treatment effect.

There are several limitations to these studies. First, these children may not have had bacterial pneumonia, as fast breathing alone (a broad diagnostic criterion) was used for diagnosis. This is supported by a study from Pakistan that analysed 1848 children with WHO-defined non-severe pneumonia who had CXRs [45]. Of these, 1519 (82%) children had normal CXRs, 263 (14%) showed radiographic evidence of pneumonia, and 66 (4%) had bronchiolitis. These findings suggest that many of these children likely had viral infections that resolved over time, regardless of antibiotic use. The WHO criteria for diagnosing CAP prioritise sensitivity over specificity to ensure early treatment and reduce mortality [46]. Therefore, these trials often lack additional clinical, radiographic, and laboratory assessments that could support a bacterial CAP diagnosis [33, 47]. Given the limited resources in these settings, this approach is understandable.

Second, children with wheezing were also included despite evidence that viral infections are independently associated with wheezing [48]. Data from seven LMICs showed viruses are the most common cause of severe pneumonia requiring hospitalisation (61%), whereas bacteria account for only 27% [35].

Third, fever was not an inclusion criterion in these studies, increasing the likelihood of wheezing illnesses being misclassified as pneumonia. In most clinical settings, fever is a key symptom of childhood CAP. A prospective study of 390 children from Brazil revealed that adding fever to the WHO criteria significantly improved the ability to differentiate pneumonia from wheezing illnesses in children, particularly by increasing the specificity of diagnosing CAP based upon clinical signs [49].

Fourth, two studies [41, 43] employed different antibiotics for the day 3 and day 5 arms, where differences in efficacy and adherence may have influenced outcomes. Finally, randomisation was conducted at enrolment, rather than at day 3, so treatment failures in the first 3 days while receiving identical therapy were also allocated to the 5-day arm, potentially diluting any beneficial effect from the longer course.

Our comments are supported by RCTs comparing antibiotics and placebo for childhood CAP. A Pakistan-based RCT randomised 873 children aged 2−59 months with WHO-defined non-severe CAP to receive either 3 days of low-dose oral amoxicillin (45 mg/kg/day) or placebo [50]. The primary outcome was treatment failure on day 3. The per-protocol analysis showed no significant intergroup difference (amoxicillin 31/431 [7.2%]; placebo 37/442 [8.3%]; OR 0.85; 95% CI 0.50, 1.43). However, conflicting evidence arose from larger studies conducted in Pakistan and Malawi that also evaluated 3 days of amoxicillin compared with placebo to treat non-severe childhood CAP [51, 52]. Both studies found that placebo was significantly inferior to amoxicillin. In the Pakistan RCT, treatment failure by day 3 occurred in 95 of 1927 (4.9%) children receiving placebo compared with 51 of 1929 (2.6%) receiving amoxicillin (absolute difference 2.3%; 95% CI 0.9, 3.7) [51]. In the Malawi RCT, the trial was stopped early after interim analysis showed the superiority of amoxicillin [52]. By day 4, treatment failure was observed in 38 of 543 (7.0%) children receiving placebo compared with 22 of 552 (4.0%) receiving amoxicillin (adjusted absolute difference 3.0%; 95% CI 0.4, 5.7). Nevertheless, most children in both RCTs had no treatment failure by day 4 (Pakistan 93%, Malawi 95%), indicating that many children recovered spontaneously without receiving antibiotics [53]. Moreover, these studies revealed a high number needed to treat (Malawi 33, Pakistan 44), highlighting the potential for over-treatment. This illustrates how the WHO clinical definition of CAP results in widespread empirical overuse and misuse of antibiotics, contributing to antibiotic resistance [54].

High-Income Countries

In Israel, 140 children aged 6−59 months with outpatient CAP, defined as alveolar pneumonia on CXR, temperature ≥38.5 °C, and a white cell count ≥15,000/mm³, were recruited [55]. There were two stages in the RCT, and all arms received amoxicillin TID (80 mg/kg/day). Stage 1 compared 3- and 10-day courses but was discontinued after higher treatment failure rates in the 3-day arm (4/10 [40%]) than in the 10-day arm (0/12 [0%]; P = 0.16). The 3-day arm was replaced with a 5-day arm (stage 2). There was no significant difference in treatment failure between the 5-day arm (0/42 [0%]) and the 10-day arm (0/56 [0%]). Overall, the treatment failure in the 3-day arm versus both the 5-day and the 10-day arms was significant (P < 0.001).

The Canadian SAFER (Short-Course Antimicrobial Therapy for Pediatric Respiratory Infections) trial recruited 281 children aged 6 months to 10 years with outpatient CAP defined as fever, clinical and radiographic signs consistent with CAP, and physician-confirmed diagnosis [56]. They were randomised to either 5 or 10 days of high-dose oral amoxicillin TID. The noninferiority margin was RD 7.5% assessed against a one-sided 97.5% confidence limit. The intention-to-treat analysis showed no difference in clinical cure at 14−21 days in the 5-day arm (108/126 [85.7%]) compared with the 10-day arm (106/126 [84.1%]; RD 0.023, 97.5% one-sided confidence limit −0.061). Baseline nasopharyngeal swabs were positive for one or more viruses in 58 of 130 (45%) children, indicating that nearly half had viral infections, and only 208 (74%) were later deemed to have radiographic changes consistent with pneumonia.

The CAP-IT (Community-Acquired Pneumonia: an RCT) trial, undertaken in the UK and Ireland, recruited 824 children aged >6 months with outpatient or inpatient CAP (defined as cough <96 h, reported fever/temperature >38 °C, and laboured/difficult breathing or focal chest signs) [57]. The children were randomised to receive low-dose (35−50 mg/kg/day BID) or high-dose (70−90 mg/kg/day BID) oral amoxicillin for 3 or 7 days. This 2 × 2 factorial noninferiority trial used a noninferiority margin of RD 8%, assessed with a one-sided 95% CI. Both groups demonstrated noninferiority for re-treatment by day 28, with no significant difference between low (51/410 [12.6%]) and high (49/404 [12.4%]) doses (difference 0.2%; one-sided 95% CI −$\infty$, 4.0) or 3 days (51/413 [12.5%]) and 7 days (49/401 [12.5%]) of treatment (difference 0.1%; one-sided 95% CI −$\infty$, 3.9). Nonetheless, in the subgroup of 223/824 (27%) hospitalised children, the re-treatment rates were 15.3% and 11.5% in the lower- and higher-dose groups, respectively (difference 3.7%; one-sided 95% CI −$\infty$, 11.4), and 15.2% and 11.3% for the 3- and 7-day treatment courses, respectively (difference 3.9%; one-sided 95% CI −$\infty$, 11.5). Neither met the noninferiority criterion.

Unlike other study designs, the CAP-IT trial focused on a distinct outcome, emphasising re-treatment rates rather than traditional endpoints such as clinical cure. Its findings are particularly relevant to the epidemiology of penicillin non-susceptibility in S. pneumoniae in the UK and Ireland, where the study was conducted [58, 59]. An analysis of 390 S. pneumoniae isolates from 1132 nasopharyngeal swabs collected from 718 children in the CAP-IT cohort showed low rates of non-susceptibility to penicillin (15.6%) and amoxicillin (2.6%) [60], but the results need to be interpreted cautiously in regions with different resistance patterns. The observed noninferiority results for both lower-dose and shorter-duration treatments raises potential concerns. Although noninferiority was demonstrated overall, the re-treatment rates in hospitalised children were notably higher in the lower-dose and shorter-duration groups, which might indicate that reduced dosing or shorter treatment may be less effective in more severe cases and in settings where penicillin non-susceptible strains of S. pneumoniae are common [16–18].

The SCOUT-CAP (Short-Course Outpatient Therapy of Community-Acquired Pneumonia) superiority trial in the USA recruited 380 children aged 6−71 months with clinician-diagnosed CAP in the outpatient setting. Children were prescribed amoxicillin, amoxicillin–clavulanate, or cefdinir for 5 days [61]. If fever, tachypnoea, and severe cough resolved by days 3−6 of antibiotic therapy, they were randomised to continue antibiotics (oral amoxicillin BID [80−100 mg/kg/day] or amoxicillin-clavulanate BID [80−100 mg/kg/day amoxicillin component] or oral cefdinir BID [12−16 mg/kg/day]) or switched to placebo for another 5 days. The primary endpoint was the end-of-treatment response adjusted for duration of antibiotic risk (RADAR), combining clinical response, symptom resolution, and adverse effects. It assumes shorter treatments are preferable when outcomes are similar. At visit 1 (days 6−10), the 5-day arm showed a 69% chance of a more favourable RADAR score (probability 0.69; 95% CI 0.63, 0.75). By visit 2 (days 19−25), the probability remained higher for the 5-day arm (0.64; 95% CI 0.57, 0.69). Despite these positive results, this pragmatic study did not specify the criteria used to diagnose CAP, which was left to the attending physician and may affect interpretation.

These studies indicate that a 3-day antibiotic treatment for children with uncomplicated CAP is controversial. The small Israeli trial found it insufficient, but the larger CAP-IT trial reported that 3 days of treatment was noninferior to 7 days in outpatient-treated children but not in those hospitalised with severe CAP. Generally, however, three studies agreed that a 5-day antibiotic treatment course for uncomplicated CAP in previously healthy children treated as outpatients was noninferior, and one study demonstrated that it was superior to a 10-day treatment after adjusting for the impact of antibiotic duration on outcomes.

High- and Middle-Income Countries

A multinational RCT was conducted involving 324 children aged 3 months to ≤5 years from communities with a high risk of chronic lung disease and who were hospitalised with CAP [62]. The study included Indigenous children from Australia and New Zealand and children from Malaysia. Eligible children had clinical signs of CAP (temperature >37.5 °C, age-adjusted tachypnoea, chest indrawing, and/or hypoxemia) and radiographic evidence of lobar/segmental airspace involvement and were hospitalised for 1−3 days of intravenous antibiotics. After improving clinically, they transitioned to 3 days of oral amoxicillin-clavulanate BID (80 mg/kg/day amoxicillin component) and were then randomised to receive either 8 days of oral amoxicillin–clavulanate (extended course 13−14 days) or 8 days of oral placebo (standard course 5−6 days). The study found no significant difference in clinical cure at 4 weeks between the extended (127/163 [77.9%]) and the standard (131/161 [81.3%]) treatment arms (relative risk [RR] 0.96; 95% CI 0.86, 1.07) [62].

Randomised Controlled Trials of Long-Term Outcomes (i.e. Chronic Respiratory Symptoms/Signs >30 days)

Currently, only one antibiotic RCT has focused on the long-term respiratory outcomes of 324 children hospitalised with CAP, whereby children were reviewed at 12 and 24 months after discharge [63]. The RCT involved the same cohort mentioned in the previous section [62] but with a different primary outcome: the composite endpoint of chronic respiratory symptoms/signs at 24 months. The endpoints included chronic cough at both 12 and 24 months, one or more subsequent hospitalised ALRI by 24 months, or persistent and/or new CXR signs at 12 months. There was no significant difference in chronic respiratory symptoms/signs between the extended-course (97/163 [60%]) and standard-course (94/161 [58%]) arms (RR 1.02; 95% CI 0.85, 1.22). The study was affected by lockdowns imposed in response to the coronavirus disease 2019 pandemic, limiting complete follow-up at 12 and 24 months to 213/324 (66%) participants and where, in the primary intention-to-treat analysis, those lost to follow-up were assumed to have chronic respiratory symptoms/signs. Nevertheless, additional analyses supported the findings.

Systematic Reviews and Meta-Analyses of Short-Term Outcomes

Currently, all systematic reviews and meta-analyses have only examined short-term CAP outcomes. A Cochrane systematic review (three RCTs, 5763 children, search date August 2010) analysed children aged 2−59 months diagnosed with WHO-defined non-severe CAP in LMICs [64]. They found that 3 days versus 5 days of treatment with the same oral antibiotic (amoxicillin or trimethoprim–sulfamethoxazole) did not show any significant difference in clinical cure rates at the end of treatment (RR 0.99; 95% CI 0.97, 1.01).

The latest systematic review of children with CAP (four RCTs, n = 1573) treated in either outpatient (n = 1350) or inpatient (n = 223) settings in HICs showed that the RD for treatment failures was 0.1% (95% CI −3.0, 2.0) [65]. The authors concluded that 3−5 days of antibiotic treatment was equally efficacious as 7−10 days in children aged ≥6 months with CAP.

Additional reviews included studies from HICs and LMICs. One (nine RCTs, n = 11,143) compared the same oral antibiotic across short- and long-course regimens, and another (16 RCTs, n=12,774) compared both similar and different antibiotics, including macrolides, across course durations [66, 67]. Both reviews concluded that shorter courses of antibiotics (3−5 days) were noninferior to longer courses (5−10 days) for treating non-severe CAP in children. Both studies also suggested that shorter antibiotic durations resulted in similar clinical outcomes, including treatment success, while potentially reducing the risk of adverse events, promoting better adherence, and minimising antibiotic resistance.

Where to From Here?

The critical question to consider before determining the antibiotic duration for treating childhood CAP is ‘whether it is bacterial pneumonia’ [68]. Predicting a bacterial, non-bacterial, or mixed aetiology based upon clinical and radiographic findings remains difficult [15]. In paediatric patients, specimens commonly collected to identify the aetiology of pneumonia include nasopharyngeal swabs/aspirates, pleural aspirates (for complicated pneumonia), blood, and urine. These samples are analysed using bacterial cultures, molecular diagnostics, or antigen detection for bacteria and viruses. These tests offer certain advantages but also have notable limitations. For example, a positive nasopharyngeal swab/aspirate only indicates the presence of organisms in the upper airway and does not confirm their role in the pneumonic process. Similarly, bacterial cultures from the blood have low diagnostic yields and are associated with high costs per case detected [30, 69, 70].

Host biomarkers are emerging as promising tools for diagnosing pneumonia. Procalcitonin and C-reactive protein remain the most used biomarkers, and interleukin-6 shows greater sensitivity in detecting localised infections (e.g. effusions) and predicting treatment failure or mortality [71]. Most studies on host biomarkers have been conducted in adults, focusing on their association with pneumonia diagnosis, severity, and mortality [71–73]. However, in children with CAP, routine biomarkers such as white cell count, C-reactive protein, and procalcitonin have shown limited utility in predicting disease severity [74]. Additionally, these biomarkers demonstrate varying sensitivity and specificity for identifying bacterial pneumonia in children [69, 75].

The ongoing challenge of distinguishing bacterial from non-bacterial pneumonia is compounded by the lack of a simple point-of-care diagnostic test [76]. However, emerging methods such as the identification of three proteins (haptoglobin, tumour necrosis factor receptor 2 or interleukin-10, and tissue inhibitor of metalloproteinases 1) have classified febrile children with respiratory symptoms into bacterial, malarial, and viral aetiologies [77]. Moreover, host gene profiles by transcriptomics show promise in distinguishing bacterial from viral infections [78], including cases with ALRI [69, 70]. These advances might enable more targeted therapies for children with bacterial CAP and optimise treatment strategies.

Limitations

There are important limitations to this review. First, we focused on β-lactam antibiotics as first-line agents for treating childhood pneumonia and did not specifically examine macrolides for treating atypical pneumonia, which in children is classically caused by M. pneumoniae. Indeed, M. pneumoniae has been identified as a cause of pneumonia in children in several aetiology studies. In the prospective EPIC (Etiology of Pneumonia in the Community) study, which included 2222 children aged <18 years hospitalised with CAP across three US hospitals, M. pneumoniae was the most frequently detected bacterial pathogen, but this involved only 178 (8%) cases [79]. Notably, M. pneumoniae was more common among children aged ≥5 years than in younger children (19% vs 3%). However, children aged <5 years worldwide have the greatest disease burden [2]. Among hospitalised children aged ≤5 years in eight LMICs, M. pneumoniae accounted for only 1.5% (13/888) of pneumonia cases [80]. Additionally, in a large cohort of 1769 children aged <5 years hospitalised with severe pneumonia without HIV infection across seven LMICs, M. pneumoniae contributed to <1% of the aetiological fraction of radiographically confirmed CAP, which was far less than S. pneumoniae [35].

M. pneumoniae is also found frequently in the upper respiratory tract of asymptomatic children [81, 82], making it difficult to reliably assign the organism as a cause of pneumonia in the absence of serology, and especially when other pathogens, such as respiratory viruses, are also present. This raises questions about whether antibiotic treatment is always necessary when M. pneumoniae is detected in a child with CAP. Two systematic reviews highlighted the lack of substantial evidence supporting antibiotic treatment for M. pneumoniae pneumonia in children [83, 84]. A study of 1418 children from the EPIC cohort found no significant difference in length of hospital stay between children receiving β-lactam monotherapy and β-lactam plus macrolide combination therapy (median 55 vs 59 h; adjusted HR 0.87; 95% CI 0.74, 1.01) [85]. Furthermore, a sub-analysis involving 110 children with M. pneumoniae also found no difference in duration of hospital stay between those treated with either β-lactam monotherapy or β-lactam plus macrolide combination therapy (adjusted HR 1.07; 95% CI 0.59, 1.96). To address this issue, a noninferiority RCT is currently under way to assess the efficacy of macrolides compared with placebo in children with M. pneumoniae CAP [86]. Meanwhile, there are concerns over widespread macrolide consumption, which is associated with not only increasing macrolide-resistant M. pneumoniae infections globally, but also resistance in other major respiratory and gastrointestinal pathogens [86–90]. Macrolides, such as azithromycin, are therefore under the ‘Watch’ category of the WHO Access, Watch, Reserve (AWaRe) antibiotic book for guiding antibiotic therapy and where indications in children for azithromycin are limited to selected enteric infections [91].

Second, the RCTs included in this review did not determine the microbiological aetiology of pneumonia. This is understandable given the challenges with identifying the cause of pneumonia at an individual level [30]. In addition, amoxicillin was the chosen antibiotic in almost all studies. This too is reasonable, as many participants may have had viral pneumonia, and amoxicillin is a relatively narrow-spectrum antibiotic that is also effective, cheap, and well-tolerated. Indeed, these properties, combined with S. pneumoniae remaining the most important cause of bacterial pneumonia, led the WHO and other producers of major guidelines to recommend amoxicillin as the oral antibiotic of choice for suspected bacterial pneumonia in children [24–26]. In contrast, Hib is now a rare cause of pneumonia in countries with a high uptake of Hib conjugate vaccines [92], and non-Hib strains, S. aureus and Streptococcus pyogenes, cause considerably less pneumonia than S. pneumoniae [2, 35, 80, 92]. Taken together, for the foreseeable future, it is unlikely that studies examining the optimal duration of antibiotic therapy for uncomplicated pneumonia in otherwise healthy children will involve second-line, broad-spectrum antibiotics. This is because many will have non-bacterial ALRIs and these antibiotics are more expensive, less well-tolerated, more likely to induce antibiotic resistance, and—in the case of trimethoprim–sulfamethoxazole—less effective than oral amoxicillin [93–95].

Third, the systematic reviews and meta-analyses on antibiotic treatment duration included studies with varying definitions of CAP and outcomes, potentially contributing to heterogeneity of results. However, one study used a random-effects model to pool the data and reported no statistical heterogeneity in the meta-analysis [67].

Fourth, we did not discuss physical examination findings for diagnosing CAP and guiding antibiotic treatment because of their limited reliability [96, 97]. Nonetheless, in resource-limited settings where CXR is unavailable, clinical indicators such as increased work of breathing and hypoxemia can help identify children with CAP [97].

Conclusion

Studies on antibiotic duration for uncomplicated pneumonia in children suffer from major limitations, the most important of which is recruiting subjects unlikely to have bacterial pneumonia, thereby biasing results in favour of the null hypothesis. Therefore, substantial uncertainties remain over the optimal duration of treatment in patients with suspected bacterial CAP where balancing effective treatment with risking emergent antibiotic resistance is required. Consequently, recommendations for antibiotic treatment duration of uncomplicated CAP in children (Table 1) should be regarded as provisional since they are based upon weak evidence. Additional research is needed to address important knowledge gaps in making an accurate diagnosis of bacterial CAP and to tailor treatment duration based upon pathogen resistance patterns in the community, illness severity, underlying comorbidities, and the child’s immunisation status. The overall goal is to ensure safe and successful antibiotic therapy and to minimise antibiotic resistance selection pressure by prescribing the shortest, most effective antibiotic course to those most likely to have bacterial CAP.

Table 1.

Provisional recommendations for duration of antibiotic treatment of uncomplicated community-acquired pneumonia in otherwise healthy children^a

Country income category	Settings	Treatment recommendations
Low- and middle-income countries	Outpatient	3 days of oral amoxicillinb for fast breathing and/or chest-indrawing community-acquired pneumoniac
High-income countries	Outpatient	5 days of oral amoxicillin,b unless Mycoplasma pneumoniae suspected
High- and middle-income countries	Inpatient	5 days of antibiotics in total for those improving within 1–3 days of parenteral antibiotics and capable of transitioning to oral amoxicillinb or amoxicillin–clavulanate in high-risk Indigenous childrend

^aDiagnosing bacterial community-acquired pneumonia remains challenging across all countries and settings, affecting the reliability of evidence and weakening recommendations

^bHigh-dose amoxicillin (90 mg/kg/day in thrice-daily dosing [maximum single dose of 1000 mg]) is the treatment of choice because it is effective against the most common bacterial pathogens, cheap, widely available, and well tolerated

^cStudies from resource-limited low- and middle-income countries support 3 days of oral amoxicillin for children aged 2−59 months with fast breathing and/or chest-indrawing pneumonia. However, many participants likely had a wheezing illness or non-bacterial pneumonia, biasing the results towards the null

^dThis includes disadvantaged Indigenous children living in communities where chronic suppurative lung disease and bronchiectasis associated with prior hospitalisations for acute lower respiratory infections, including pneumonia, are common

Supplementary Information

Below is the link to the electronic supplementary material.

Declarations

Conflict of Interest

ABC, GBM, STY, and KG have received grants from the Australian National Health and Medical Research Council (NHMRC) and NHMRC-managed grants (Medical Research Futures Fund). ABC is also an independent data management committee member for clinical trials for Moderna (COVID-19 and Epstein-Barr virus vaccines) and of an unlicensed vaccine (GlaxoSmithKline) and monoclonal antibody (AstraZeneca) and has received fees to the institution for consulting on the study designs for Zambon and Boehringer Ingelheim, airfares for travel from the European Respiratory Society and Boehringer Ingelheim, and personal fees for being an author of two UpToDate chapters that are outside the submitted work. HCK and SMF have no conflicts of interest to disclose.

Funding

Open access funding enabled and organized by CAUL and its Member Institutions. No funds, grants, or other support were received during the preparation of this manuscript. HCK is supported by the Malaysian Ministry of Health and a Charles Darwin International PhD Scholars scholarship, and ABC is supported by an NHMRC L3 fellowship (grant 2025379).

Ethics Approval

Not applicable.

Consent to Participate

Not applicable.

Consent for Publication

Not applicable.

Availability of Data and Material

Not applicable.

Code Availability

Not applicable.

Author Contributions

HCK wrote the first draft of the manuscript, with revisions from KG and ABC. All authors contributed to editing the manuscript and approved the final version.