Pneumonia is the leading single cause of mortality in children aged less than five years, with an estimated incidence of 0.29 and 0.05 episodes per child-year in low-income and high-income countries respectively. It is estimated that a total of around 156 million new episodes occur each year and most of these occur in India (43 million), China (21 million), Pakistan (10 million) and Bangladesh, Indonesia and Nigeria (six million each) (Rudan 2008). Pneumonia is responsible for about two million deaths each year in children below five years of age and these occur mainly in the African and South-East Asian regions. Pneumonia contributes to about one-fifth (19%) of all deaths in children aged less than five years, of which more than 70% take place in sub-Saharan Africa and South-East Asia (Rudan 2008). To reduce the infant and under-five child mortality, it is important to reduce mortality due to pneumonia by appropriate intervention in the form of antibiotics. Selection of first-line antibiotics for empirical treatment of pneumonia is crucial for office practice as well as public health.
Description of the condition
Pneumonia is defined as infection of lung parenchyma (alveoli) by microbial agents. It is difficult to identify the causative organism in most cases of pneumonia. The methods used for identification of the aetiologic agents include blood culture, lung puncture, nasopharyngeal aspiration and immune assays of blood and urine tests. Lung puncture is an invasive procedure associated with significant morbidity and hence cannot be performed routinely in most cases. The yield from blood cultures is too low (5% to 15% for bacterial pathogens) to be relied upon (MacCracken 2000).There are few studies that document the aetiology of pneumonia in children below five years of age from low-income countries. Most studies carried out blood cultures for bacterial aetiology of pneumonia. Some studies carried out nasopharyngeal aspirates and identification of atypical organisms. A review of 14 studies involving 1096 lung aspirates taken from hospitalised children prior to administration of antibiotics reported bacterial pathogens in 62% of cases (Berman 1990). In 27% of patients, the common bacterial pathogens identified were Streptococcus pneumoniae (S. pneumoniae) and Haemophilus influenzae (H. influenzae) (Berman 1990). Studies using nasopharyngeal aspirates for identification of viral agents suggest that about 40% of pneumonia in children below five years of age is caused by viral agents, with the commonest viral pathogen being respiratory syncytial virus (Maitreyi 2000). In infants under three months of age, common pathogens include S. pneumoniae, H. influenzae, gram-negative bacilli and Staphylococcus (WHOYISG 1999). The causative organisms are different in high-income countries and include more viral and atypical organisms (Gendrel 1997; Ishiwada 1993; Numazaki 2004; Wubbel 1999). Therefore, treatment regimens may be different in high-income and low-income countries.
Description of the intervention
Administration of appropriate antibiotics at an early stage of pneumonia improves the outcome of the illness, particularly when the causative agent is bacterial. The World Health Organization (WHO) has provided guidelines for early diagnosis and assessment of the severity of pneumonia on the basis of clinical features (WHOYISG 1999) and suggests administration of co-trimoxazole as a first-line drug. The commonly used antibiotics for community-acquired pneumonia (CAP) include co-trimoxazole, amoxycillin, oral cephalosporins and macrolide drugs. Despite evidence of rising bacterial resistance to co-trimoxazole (IBIS 1999; Timothy 1993), studies conducted in the same time period showed good clinical efficacy of oral co-trimoxazole for non-severe pneumonia (Awasthi 2008; Rasmussen 1997; Straus 1998). However, one study reported a doubling of clinical failure rates with co-trimoxazole treatment when compared to treatment with amoxycillin in severe and radiologically confirmed pneumonia (Straus 1998). A meta-analysis of all the trials on pneumonia based on the case-management approach proposed by WHO (identification of pneumonia on clinical symptoms/signs and administration of empirical antimicrobial agents) has found reduction in mortality as well as pneumonia-related mortality (Sazawal 2003). To meet the public health goal of reducing child mortality due to pneumonia, empirical antibiotic administration is relied upon in most instances. This is necessary in view of the inability of most commonly available laboratory tests to identify causative pathogens.
Why it is important to do this review
Empirical antibiotic administration is the mainstay of treatment of pneumonia in children. Administration of the most appropriate antibiotic as first-line medicine may improve outcome of pneumonia. There are multiple antibiotics prescribed for treatment of pneumonia, therefore it is important to know which work best for pneumonia in children. The last review of all available randomised controlled trials (RCTs) on antibiotics used for pneumonia in children was published in 2006 (Kabra 2006). Since then, several new studies (Asghar 2008; Atkinson 2007; Aurangzeb 2003; Bansal 2006; Bradley 2007; Esposito 2005; Hasali 2005; Hazir 2008; Lee 2008; Lu 2006) have been published. It is therefore important to update the information by including all the new clinical trials.
To identify effective antibiotic drug therapies for community-acquired pneumonia (CAP) in children by comparing various antibiotics.
Criteria for considering studies for this review
Types of studies
Randomised controlled trials (RCTs) comparing antibiotics for community-acquired pneumonia (CAP) in children. We considered only those studies using the case definition of pneumonia (as given by the World Health Organization (WHO)) or radiologically-confirmed pneumonia in this review.
Types of participants
We included children under 18 years of age with CAP treated in a hospital or community setting. We excluded studies describing pneumonia post-hospitalisation in immunocompromised patients (for example, following surgical procedures).
Types of interventions
We compared any intervention with antibiotics (administered by intravenous route, intramuscular route, or orally) with another antibiotic for the treatment of CAP.
Types of outcome measures
- Clinical cure. Definition of clinical cure is symptomatic and clinical recovery by the end of treatment.
- Treatment failure rates. Definition of treatment failure is the presence of any of the following: development of chest in-drawing, convulsions, drowsiness or inability to drink at any time, respiratory rate above the age-specific cut-off point on completion of treatment, or oxygen saturation of less than 90% (measured by pulse oximetry) after completion of the treatment. Loss to follow up or withdrawal from the study at any time after recruitment was taken as failure in the analysis.
The clinically relevant outcome measures were as follows.
- Relapse rate: defined as children declared 'cured', but developing recurrence of disease at follow up in a defined period.
- Hospitalisation rate (in outpatient studies only). Defined as need for hospitalisation in children who were getting treatment on an ambulatory (outpatient) basis.
- Length of hospital stay: duration of total hospital stay (from day of admission to discharge) in days.
- Need for change in antibiotics: children required change in antibiotics from the primary regimen.
- Additional interventions used: any additional intervention in the form of mechanical ventilation, steroids, vaso-pressure agents, etc.
- Mortality rate.
Search methods for identification of studies
We retrieved studies through a search strategy which included cross-referencing. We checked the cross-references of all the studies by hand.
We searched the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2009, issue 2), which contains the Acute Respiratory Infections Group's Specialised Register, MEDLINE (1966 to September 2009) and EMBASE (1990 to September 2009). There were no language or publication restrictions. We combined the MEDLINE search with the Cochrane Highly Sensitive Search Strategy for identifying randomised trials in MEDLINE: sensitivity- and precision-maximising version (2008 revision); Ovid format (Lefebvre 2008). See Appendix 1 for the EMBASE search strategy.
1 exp PNEUMONIA/
4 exp Anti-Bacterial Agents/
7 exp CHILD/
8 exp INFANT/
9 (children or infant$ or pediatric or paediatric)
11 3 and 6 and 10
Searching other resources
We also searched bibliographies of selected articles to identify any additional trials not recovered by the electronic searches.
Data collection and analysis
Selection of studies
Two review authors (SKK, RL) independently selected potentially relevant studies based on their title and abstract. The complete texts of these studies were retrieved electronically or by contacting the trial authors. Two review authors (SKK, RL) independently reviewed the results for inclusion.
Data extraction and management
All the relevant studies were masked for authors' names and institutions, the location of the study, reference lists and any other potential identifiers. The papers were then given a serial number by a person who was not involved in the review. Two review authors (SKK, RL) independently reviewed the results for inclusion in the analysis. Differences about study quality were resolved through discussion. We recorded data on a pre-structured data extraction form. We assessed publication bias using the Cochrane Collaboration's 'Risk of bias' tool. Before combining the studies for each of the outcome variables, we carried out assessment of heterogeneity with Breslow's test of homogeneity using the Review Manager (RevMan) software (version 5.0) (RevMan 2008). We performed sensitivity analysis to check the importance of each study in order to see the effect of inclusion and exclusion criteria. We computed both the effect size and summary measures with 95% confidence intervals (CI) using RevMan software. For all the outcome variables, we used a random-effects model to combine the study results.
We collected data on the primary outcome (cure rate/failure rate) and secondary outcomes (relapse rate, rate of hospitalisation and complications, need for change in antibiotics, need for additional interventions and mortality). When available, we also recorded additional data on potential confounders such as underlying disease, prior antibiotic therapy and nutritional status.
We did multiple analyses, firstly on studies comparing the same antibiotics. We also attempted to perform indirect comparisons of various drugs when studies with direct comparisons were not available. For example, we compared antibiotics A and C when a comparison of antibiotics A and B was available and likewise a separate comparison between antibiotics B and C. We only did this type of comparison if the inclusion and exclusion criteria of these studies, the dose and duration of the common intervention (antibiotic B), baseline characteristics and the outcomes assessed were similar (Bucher 1997).
Assessment of risk of bias in included studies
We assessed risk of bias in all included studies using the Cochrane Collaboration's 'Risk of bias' tool (Higgins 2008):
1. Sequence generation: assessed as yes, no or unclear
Yes: when the study described the method used to generate the allocation sequence in sufficient detail.
No: sequence not generated.
Unclear: when it was not described or incompletely described.
2. Allocation concealment: assessed as yes, no or unclear
Yes: when the study described the method used to conceal the allocation sequence in sufficient detail.
No: described details where allocation concealment was not done.
Unclear: when it was not described or incompletely described.
3. Blinding of participants, personnel and outcome assessors: assessed as yes, no or unclear
Yes: when it was a double-blind study.
No: when it was an unblinded study.
Unclear: not clearly described.
4. Incomplete outcome data: assessed as yes, unclear
Yes: describe the completeness of outcome data for each main outcome, including attrition and exclusions from the analysis.
Unclear: either not described or incompletely described.
5. Free of selective outcome reporting: assessed as yes, no or unclear
Yes: results of study free of selective reporting. Details of all the patients enrolled in the study are included in the paper.
No: details of all the enrolled patients not given in the paper.
Unclear: details of all the enrolled patients incompletely described.
6. Other sources of bias
Among the other sources of potential bias considered was funding agencies and their role in the study. We recorded funding agencies as governmental agencies, universities and research organisations or pharmaceutical companies. We considered studies supported by pharmaceutical companies to be unclear unless the study defined the role of the pharmaceutical companies. We also considered studies not mentioning the source of funding as unclear under this heading.
Assessment of heterogeneity
For each of the outcome variables, we carried out assessment of heterogeneity with Breslow's test of homogeneity using RevMan software (as described in data extraction and analysis section).
Assessment of reporting biases
Before combining the study results we checked for publication bias by using a funnel plot. For each of the outcome variables (cure rate, failure rate, relapse rate, rate of hospitalisation, the complications needed for change in antibiotics and mortality rate) we used a two-by-two table for each study and performed Breslow's test of homogeneity to determine variation in study results.
Description of studies
Results of the search
Two review authors (SKK, RL) screened the article titles. Forty-four studies were short-listed as potential randomised controlled trials to be included and we attempted to collect the full-text articles; we obtained the full text for 43. These papers were blinded by a third person who was not involved in the review. Two review authors (SKK, RL) independently extracted data by using a pre-designed data extraction form; the extracted data matched completely.
We identified 27 studies for inclusion, with the following comparisons.
- Clarithromycin with erythromycin: one study (Block 1995), involving 357 children below 15 years of age with clinical or radiographically diagnosed pneumonia treated on an ambulatory basis.
- Chloramphenicol with penicillin and gentamycin together: one study (Duke 2002), involving 1116 children aged one month to five years.
- Amoxycillin with procaine penicillin: one study (Tsarouhas 1998), involving 170 children aged six months to 18 years.
- Ampicillin with chloramphenicol plus penicillin: one study (Deivanayagam 1996), involving 115 children aged five months to four years.
- Co-trimoxazole with single-dose procaine penicillin followed by oral ampicillin: one study (Campbell 1988), involving 134 children aged below five years.
- Co-trimoxazole with chloramphenicol: one study (Mulholland 1995), involving 111 children aged under five years.
- Cefpodoxime with co-amoxyclavulanic acid: one study (Klein 1995), involving 348 children aged three months to 11.5 years.
- Azithromycin with amoxycillin: one study (Kogan 2003), involving 47 children aged one month to 14 years.
- Amoxycillin with co-amoxyclavulanic acid: one study (Jibril 1989), involving 100 children aged two months to 12 years.
- Chloramphenicol in addition to penicillin with ceftriaxone: one study (Cetinkaya 2004), involving 97 children aged between two to 24 months admitted to hospital with severe pneumonia.
- Levofloxacin and comparator (co-amoxyclavulanic acid or ceftriaxone): one study (Bradley 2007) involving 709 children aged 0.5 to 16 years of age with community-acquired pneumonia treated in hospital or ambulatory care.
- Parenteral ampicillin followed by oral amoxycillin with home-based oral amoxycillin: one study (Hazir 2008) involving 2037 children between three months to 59 months of age with WHO-defined severe pneumonia.
- Chloramphenicol with ampicillin and gentamycin: one study (Asghar 2008), involving 958 children between two to 59 months with very severe pneumonia.
- Penicillin and gentamicin with co-amoxyclavulanic acid (Bansal 2006), involving 71 children with severe and very severe pneumonia between two months to 59 months of age.
- Co-amoxyclavulanic acid with cefuroxime or clarithromycin: one study (Aurangzeb 2003), involving 126 children between two to 72 months of age.
We excluded a total of 17 studies.
- One study did not provide separate data for children (Sanchez 1998).
- One studied only sequential antibiotic use (Al-Eiden 1999).
- One compared azithromycin with symptomatic treatment for recurrent respiratory tract infection only (Esposito 2005).
- Full text article could not be obtained for one study (Lu 2006).
- One study (Lee 2008) was excluded because the outcome was not in the form of cure or failure rates.
Risk of bias in included studies
Details of sequence generation were described in 16 studies (Addo-Yobo 2004; Asghar 2008; Atkinson 2007; Awasthi 2008; Bansal 2006; Camargos 1997; CATCHUP 2002; Cetinkaya 2004; Deivanayagam 1996; Duke 2002; Hazir 2008; Jibril 1989; Keeley 1990; Mulholland 1995; Roord 1996; Shann 1985), were not clear in nine studies (Aurangzeb 2003; Block 1995; Bradley 2007; Campbell 1988; Harris 1998; Klein 1995; Straus 1998; Tsarouhas 1998; Wubbel 1999) and sequence was not generated in two studies (Kogan 2003; Sidal 1994).
Allocation concealment was adequate in 16 studies (Addo-Yobo 2004; Asghar 2008; Atkinson 2007; Awasthi 2008; Bansal 2006; Camargos 1997; CATCHUP 2002; Cetinkaya 2004; Deivanayagam 1996; Duke 2002; Harris 1998; Hazir 2008; Keeley 1990; Mulholland 1995; Shann 1985; Tsarouhas 1998), it was unclear in eight studies (Aurangzeb 2003; Block 1995; Bradley 2007; Campbell 1988; Jibril 1989; Klein 1995; Straus 1998; Wubbel 1999) and no concealment was done in three studies (Kogan 2003; Roord 1996; Sidal 1994) (see Table 1).
Incomplete outcome data
Data was fully detailed in 18 studies (Addo-Yobo 2004; Asghar 2008; Atkinson 2007; Aurangzeb 2003; Awasthi 2008; Bansal 2006; Block 1995; Camargos 1997; CATCHUP 2002; Cetinkaya 2004; Duke 2002; Hazir 2008; Kogan 2003 Mulholland 1995; Roord 1996; Straus 1998; Tsarouhas 1998; Wubbel 1999) and in the remaining studies details of attrition and exclusions from the analysis were unavailable.
Selective reporting of data was unclear in 12 studies (Atkinson 2007; Aurangzeb 2003; Bradley 2007; Campbell 1988; Deivanayagam 1996; Harris 1998; Jibril 1989; Keeley 1990; Klein 1995; Shann 1985; Sidal 1994; Wubbel 1999). The rest of the studies scored 'yes' for being free from selective reporting.
Other potential sources of bias
The source of funding was not mentioned in eight studies (Aurangzeb 2003; Bansal 2006; Cetinkaya 2004; Deivanayagam 1996; Jibril 1989; Klein 1995; Sidal 1994; Tsarouhas 1998). Six studies were funded by pharmaceutical companies (Block 1995; Bradley 2007; Duke 2002; Harris 1998; Roord 1996; Wubbel 1999). Twelve studies were supported by the WHO, Medical Research Council or universities (Addo-Yobo 2004; Asghar 2008; Atkinson 2007; Awasthi 2008; Camargos 1997; Campbell 1988; Hazir 2008; Keeley 1990; Mulholland 1995; Shann 1985; Sidal 1994; Straus 1998). One study (CATCHUP 2002) was supported by the WHO in addition to pharmaceutical companies. Information on clearance by Ethics Committees or Institutional Review Boards was available for all except four studies (Aurangzeb 2003; Jibril 1989; Keeley 1990; Sidal 1994).
Effects of interventions
Studies comparing ambulatory treatment of non-severe pneumonia
Azithromycin versus erythromycin (Analysis 1)
Four studies (Harris 1998; Kogan 2003; Roord 1996; Wubbel 1999) compared erythromycin with azithromycin and enrolled 623 children. One study (Harris 1998) was double-blinded with adequate allocation concealment and three studies (Kogan 2003; Roord 1996; Wubbel 1999) were unblinded and did not have adequate allocation concealment. Information on the presence of wheezing was available in two studies (Harris 1998; Kogan 2003): 104 out of 318 (33%) children experienced wheezing in the azithromycin group, while 62 out of 161 (39%) in the erythromycin group experienced wheezing. The failure rates in the azithromycin and erythromycin groups were six out of 236 (2.5%) and seven out of 156 (4.4%), respectively (OR 0.57; 95% CI 0.14 to 2.33) There were no significant side effects in either group. Three studies reported data on aetiologic organisms separately for each of the two treatment groups (Harris 1998; Kogan 2003; Roord 1996); there were 234 organisms identified in the azithromycin group and 135 in the erythromycin group (Roord 1996). The distribution of different organisms was similar in the two groups. There were 24 organisms identified in the fourth study (Wubbel 1999) in 59 participants tested.
Clarithromycin versus erythromycin (Analysis 2)
One study (Block 1995) compared erythromycin and clarithromycin; 234 children below 15 years of age with clinical or radiographically diagnosed pneumonia were treated on an ambulatory basis. The trial was single-blinded and allocation concealment was unclear. The following outcomes were similar between the two groups: cure rate (OR 1.61; 95% CI 0.84 to 3.08), clinical success rate (OR 1.92; 95% CI 0.45 to 8.23), failure rate (OR 0.52; 95% CI 0.12 to 2.23), relapse rate (OR 0.17; 95% CI 0.02 to 1.45) and adverse events (OR 1.07; 95% CI 0.6 to 1.90). Resolution of pneumonia (diagnosed radiologically) was more frequent in the clarithromycin group as compared to the erythromycin group (OR 2.51; 95% CI 1.02 to 6.16). However, there were no differences in the improvement rates (OR 3.55; 95% CI 0.7 to 18.04) or decline rates (OR 0.34; 95% CI 0.06 to 1.80), both of which were established with radiological evidence.
Azithromycin versus co-amoxyclavulanic acid (Analysis 3)
Two studies (Harris 1998; Wubbel 1999) compared these two drugs in 283 children below five years of age. One study (Harris 1998) was double-blinded and allocation concealment was adequate while the other study (Wubbel 1999) was unblinded with inadequate allocation concealment. The cure rates (available for one study) (OR 1.02; 95% CI 0.54 to 1.95), failure rates (available for both studies) (OR 1.21; 95% CI 0.42 to 3.53) and improvement rates (OR 0.85; 95% CI 0.43 to 1.71) were similar in the two groups. There were fewer side effects reported in the azithromycin group (OR 0.15; 95% CI 0.04 to 0.61). The organisms isolated were S. pneumoniae in 28 children, H. influenzae in one, Mycoplasma pneumoniae (M. pneumoniae) in 36 and Chlamydia pneumoniae (C. pneumoniae) in 20. The separate data for isolation of organisms in the two groups was available in one study only (Harris 1998). The organisms isolated in this study (Harris 1998) were S. pneumoniae and H. influenzae in one patient each in the azithromycin group. Investigations for mycoplasma were positive in 21 out of the 129 children (16%) tested in the azithromycin group and 9 out of the 66 children (14%) tested in the co-amoxyclavulanic acid group. Investigations for C. pneumoniae were positive in 13 out of the 129 children (10%) tested in the azithromycin group and four out of the 66 children (6%) tested in the co-amoxyclavulanic acid group.
Azithromycin versus amoxycillin (Analysis 4)
One study involving 47 children aged between one month and 14 years with classical pneumonia compared these two drugs (Kogan 2003). Children treated with azithromycin were older than those treated with amoxycillin (OR 58.1; 95% CI 35.59, 80.61). The study was unblinded and allocation concealment was also inadequate. All children recovered at the end of treatment in both the groups. There were 19 organisms identified in the 47 children tested (10 in the azithromycin group and nine in the amoxycillin group). The identification rates were similar in the two groups. Organisms included M. pneumoniae (in five and three children for the azithromycin and amoxycillin groups, respectively), S. pneumoniae (in four and three respectively) and others (in one and three, respectively).
Amoxycillin versus procaine penicillin (Analysis 5)
One study involved 170 children aged six months to 18 years was identified (Tsarouhas 1998). The study was unblinded but allocation concealment was adequate. The age distribution in the two groups was comparable. The failure rates were similar in the two groups (OR 0.75; 95% CI 0.17 to 3.25).
Co-amoxyclavulanic acid versus amoxycillin (Analysis 6)
One study involving 100 children between two and 12 years of age. It was an open-label study on children suffering from clinically diagnosed bacterial pneumonia (Jibril 1989). The study was unblinded and allocation concealment was also inadequate. Age and sex distribution, presence of wheeze and mean weight in the two groups were comparable. Cure rate was better with co-amoxyclavulanic acid (OR 10.44; 95% CI 2.85 to 38.21).
Co-trimoxazole versus amoxycillin (Analysis 7)
Three multicentre studies (Awasthi 2008; CATCHUP 2002; Straus 1998) involving 3952 children (2067 in the co-trimoxazole group and 1885 in the amoxycillin group) between two months and 59 months of age have compared co-trimoxazole and amoxycillin. The diagnosis of pneumonia was based on clinical criteria. Two studies (CATCHUP 2002; Straus 1998) were double-blinded and allocation concealment was adequate. A third study (Awasthi 2008) was open-label and cluster-randomisation was done (the randomisation unit was Primary Health Centre) and in this study assessment of the primary outcome of treatment failure was done on day four for the amoxycillin group and day six for the co-trimoxazole group. All studies included children with non-severe pneumonia; one study (Straus 1998) also included 301 children with severe pneumonia. In pooled data the failure rate in non-severe pneumonia was similar in the two groups (OR 0.92; 95% CI 0.58 to 1.47). The cure rate could be extracted in two studies (Awasthi 2008; CATCHUP 2002) and it was not different in either treatment group (OR 1.12; 95% CI 0.61 to 2.03). Loss to follow up was comparable in the two groups (OR 0.88; 95% CI 0.67 to 1.16). There were only two deaths in both the groups. The organisms isolated from blood cultures were H. influenzae in 79 children (52 in the co-trimoxazole group and 27 in the amoxycillin group) and S. pneumoniae in 49 children (36 in the co-trimoxazole group and 13 in the amoxycillin group); the distribution was similar in the two groups. In view of the difference in the time of assessment for primary outcome in one study (Awasthi 2008) we performed analysis for failure rates in non-severe pneumonia after excluding this study. The results did not alter significantly; failure rates in the two groups were similar (OR 1.19; 95% CI 0.92 to 1.53). Failure rates in severe pneumonia available in one study was similar in the two groups (OR 1.71; 95% CI 0.94 to 3.11).
Co-trimoxazole versus procaine penicillin (Analysis 8)
Two studies (Keeley 1990; Sidal 1994) enrolled 723 children between three months and 12 years of age. Both studies were unblinded and allocation concealment was adequate in one study (Keeley 1990). The cure rate was similar in the two groups (OR 1.58; 95% CI 0.26 to 9.69). Rate of hospitalisation was available in only one study and was similar in the two groups (OR 2.52; 95% CI 0.88 to 7.25). There was only one death.
Co-trimoxazole versus single-dose procaine penicillin followed by oral ampicillin for five days (Analysis 9)
One study was included that had enrolled 134 children below five years of age with severe pneumonia as defined by WHO criteria (Campbell 1988). The study was unblinded and allocation concealment was not clearly stated. The cure rates (OR 1.15; 95% CI 0.36 to 3.61), hospitalisation rates (OR 1.57; 95% CI 0.25 to 9.72) and death rates were similar for the two groups.
Cefpodoxime versus co-amoxyclavulanic acid (Analysis 10)
One multicentre study (Klein 1995) enrolled 348 children between three months and 11.5 years of age. The study was unblinded and allocation concealment was inadequate. The age distribution in the two groups was comparable. The response rate at the end of 10 days of treatment was comparable in the two groups (OR 0.69; 95% CI 0.18 to 2.60). Organisms were isolated in 59 cases. These organisms were H. influenzae in 28 participants (47.5%), S. pneumoniae in 14 (23%), M. catarrhalis in seven (11.9%) and H. parainfluenzae in four (6.8). There was no significant difference in the bacteriologic efficacy of either group (100% versus 96.4%).
Studies comparing treatment of hospitalised children with severe/very severe pneumonia
Chloramphenicol versus penicillin plus gentamycin (Analysis 11)
One multicentre study including 1116 children aged between one month and five years compared chloramphenicol with penicillin and gentamycin. This was an open-label RCT in children with severe pneumonia that was carried out in Papua New Guinea (Duke 2002). Allocation concealment was adequate. There was no significant difference between the two groups in positive cultures, children who had received antibiotics earlier and loss to follow up. Need for change in antibiotics (OR 0.80; 95% CI 0.54 to 1.18), death rates (OR 1.25; 95% CI 0.76 to 2.07) and adverse events (OR 1.26; 95% CI 0.96 to 1.66) were similar in the two groups. However, re-admission rates before 30 days favoured the penicillin-gentamycin combination over chloramphenicol (OR 1.61; 95% CI 1.02 to 2.55). Bacterial pathogens were identified in 144 children (67 in children receiving chloramphenicol and 77 in the other group). Isolation rates or sensitivity of the organism and failure rates did not differ between the two groups.
Chloramphenicol with ampicillin and gentamycin (Analysis 12)
One multicentre study was identified; this study enrolled 958 children who were hospitalised with WHO defined very severe pneumonia (Asghar 2008). The study was unblinded and allocation concealment was adequate. Mean age, proportion of boys and number of children who had received antibiotics before enrolment were comparable in the two groups. Failure rates on day five (OR 1.51; 95% CI 1.04 to 2.19), day 10 (OR 1.46; 95% CI 1.04 to 2.06) and day 21 (OR 1.43; 95% CI 1.03 to 1.98) were significantly higher in those receiving chloramphenicol as compared to ampicillin and gentamycin. Death rates were higher in those receiving chloramphenicol (OR 1.65; 95% CI 0.99 to 2.77).
Chloramphenicol plus penicillin versus ceftriaxone (Analysis 13)
One double-blind study fulfilled the inclusion criteria; the study enrolled 97 children between 2 and 24 months of age diagnosed with severe CAP with probable bacterial aetiology (Cetinkaya 2004). Allocation concealment was adequate. Ages in the two groups were comparable (details not available). Cure rates in the two groups were similar (OR 1.36; 95% CI 0.47 to 3.93).
Chloramphenicol alone versus chloramphenicol plus penicillin (Analysis 14)
One study (Shann 1985) from Papua New Guinea involved 748 hospitalised children (age not clear) with severe pneumonia. The study was unblinded but allocation concealment was adequate. Need for change in antibiotics (OR 0.49; 95% CI 0.12 to 1.97), loss to follow up (OR 1.11; 95% CI 0.80, 1.53) and deaths rates (OR 0.73; 95% CI 0.48 to 1.09) were comparable in the two groups.
Ampicillin alone versus penicillin with chloramphenicol (Analysis 15 )
One trial involving 115 children between five months and four years of age was identified (Deivanayagam 1996). The study was unblinded and allocation concealment was adequate. Age and sex distribution and proportion of children with severe malnutrition were comparable in the two groups. The cure rates (OR 0.48; 95% CI 0.15 to 1.51) and duration of hospitalisation were similar in the two groups (weighted mean difference (WMD) 0.1; 95% CI -1.13 to 0.93).
Benzathine penicillin versus procaine penicillin (Analysis 16)
Two studies fulfilled the inclusion criteria; one which included 176 children between two and12 years of age with chest X-ray films showing lobar consolidation or infiltration (presumed streptococcal infection) (Camargos 1997) and another study of 105 children between three months and 14 years of age (Sidal 1994). Both studies were unblinded and allocation concealment was adequate in one (Camargos 1997). Cure rates were not significantly different in the two groups (OR 0.53; 95% CI 0.27 to 1.01). Failure rates were also similar between the groups (OR 3.17; 95% CI 0.9 to 11.11). Bacterial pathogens were identified in only one study. The isolation rate for S. pneumoniae was six out of 90 blood cultures performed (four patients in the benzathine group and two in the procaine penicillin group). The clinical outcome did not differ in relation to the organism identified.
Amoxycillin versus penicillin (Analysis 17)
Two multicentre non-blinded studies were identified; these enrolled 1702 children between three months and 59 months of age, suffering from severe pneumonia (diagnosed on the basis of WHO criteria) (Addo-Yobo 2004) and 203 children with radiographically confirmed pneumonia (Atkinson 2007). The studies were unblinded and allocation concealment was adequate. The second study (Atkinson 2007) measured outcome as time from randomisation until the temperature was < 38 degrees celsius for 24 hours and oxygen requirement had ceased. However, it provided data on need for change of antibiotics due to worsening of respiratory/radiological findings. For the purposes of this analysis we considered them as failure on day five. Age, sex, severe malnutrition, breast feeding and the number of children who had received antibiotics in the last week were similar in both the groups. The failure rates measured at 48 hours (OR 1.03; 95% CI 0.81 to 1.31), five days (OR 1.15; 95% CI 0.58, 2.30) and 14 days (OR 1.04; 95% CI 0.84 to 1.29) were similar in both groups. There were seven deaths in the group receiving penicillin in one study ( Addo-Yobo 2004) while no deaths were observed in the other study (Atkinson 2007).
Amoxycillin with intravenous (IV) ampicillin (Analysis 18)
One non-blinded study, involving 237 children between two and 59 months of age with severe pneumonia was identified (Hazir 2008). Allocation concealment was adequate. Number of infants in each group, sex distribution and presence of wheeze were comparable in the two groups. Failure rates (OR 0.86; 95% CI 0.63 to 1.19), relapse rates (OR 0.78; 95% CI 0.46 to 1.33) and death rates (OR 0.25; 95% CI 0.03 to 2.21) were similar in the two groups.
Amoxycillin with cefuroxime (Analysis 19)
One randomised, non-blinded controlled study was identified; this included 83 children with non-severe and severe pneumonia (Aurangzeb 2003). Allocation concealment was unclear. Baseline data in the form of mean age and proportion of boys were similar in the two groups. Cure rates (OR 2.05; 95% CI 0.18 to 23.51) and failure rates (OR 0.49; 95% CI 0.04 to 5.59) were similar in the two groups.
Amoxycillin with clarithromycin (Analysis 20)
One randomised, non-blinded controlled study compared these two drugs; 85 children with non-severe and severe pneumonia were enrolled (Aurangzeb 2003). The sequence generation and allocation concealment in the study is not clear. Baseline data in form of mean age and proportion of boys were similar in the two groups. Cure rates (OR 1.05; 95% CI 0.06 to 17.40) and failure rates (OR 0.95; 95% CI 0.06 to 15.74) were similar in the two groups.
Penicillin and gentamycin with co-amoxyclavulanic acid (Analysis 21)
One study involving 71 children between two months and 59 months of age with very severe pneumonia fulfilled the inclusion criteria (Bansal 2006). The study was non-blinded and allocation concealment was adequate. Baseline characteristics, including number of infants and sex distribution, were comparable. Failure rates in the two groups were similar (OR 0.86; 95% CI 0.05 to14.39).
Levofloxacin with comparator group (Analysis 22)
One non-blinded study, involving 709 children below 16 years of age, compared oral levofloxacin with either ceftriaxone or co-amoxyclavulanic acid (Bradley 2007). Sequence generation and allocation concealment is not clear from the study. The mean age, sex and number who received antibiotics before enrolment were comparable in the two groups. Cure rates were similar in the two groups (OR 1.05; 95% CI 0.46 to 2.42).
Cefuroxime with clarithromycin (Analysis 23)
One randomised, non-blinded, controlled study involving 85 children with non-severe and severe pneumonia was identified (Aurangzeb 2003). Allocation concealment was unclear. Baseline data in the form of mean age and proportion of boys were similar in the two groups. Cure rates (OR 0.51; 95% CI 0.04 to 5.89) and failure rates (OR 2.05; 95% CI 0.18 to 23.51) were similar in the two groups.
Co-trimoxazole versus chloramphenicol (Analysis 24)
One double-blind study involving 111 malnourished children under five years of age fulfilled the inclusion criteria for this review (Mulholland 1995). Allocation concealment was adequate. The age and sex distribution, nutritional status, children with wheezing and numbers excluded were similar in the two groups. Cure rates (OR 1.06; 95% CI 0.47 to 2.40), failure rates (OR 1.03; 95% CI 0.45 to 2.33), number of patients requiring a change in antibiotics (OR 1.42; 95% CI 0.46 to 4.40), relapse rates (OR 1.02; 95% CI 0.24 to 4.30) and death rates (OR 2.21; 95% CI 0.63 to 7.83) were similar in the two groups.
Oral treatment of severe pneumonia with parenteral treatment (Analysis 25)
There were three studies that compared oral amoxycillin with injectable penicillin in two studies (Addo-Yobo 2004; Atkinson 2007) or ampicillin in one study (Hazir 2008). A total of 3942 children were enrolled. Two studies (Addo-Yobo 2004; Hazir 2008) enrolled 3739 children below five years of age, while the third study (Atkinson 2007) enrolled a total of 203 children; of these 36 were above 60 months of age. The baseline characteristics in the form of age and sex distribution in the two groups and proportion of children who had received antibiotics before enrolment were comparable in the two groups. Failure rates were available in all three studies and were similar in the two groups (OR 0.95; 95% CI 0.78 to 1.15). Death rates were significantly more in those who received injectable treatment (OR 0.15; 95% CI 0.03 to 0.87). There were no deaths in one study (Atkinson 2007), seven deaths in another study (but all happened in those getting injectable medications (Addo-Yobo 2004)) and five deaths in the third study (one in the oral group and four in the injectable group) (Hazir 2008). Reanalysis after removing one study (Addo-Yobo 2004) with seven deaths in only one group suggests no significant difference in the two groups (OR 0.87; 95% CI 0.65 to 1.15). Relapse rates could be derived from two studies (Atkinson 2007; Hazir 2008) and were similar in the two groups (OR 1.28; 95% CI 0.34 to 4.82).
Identification of aetiological agents
Out of 27 studies reviewed, attempts were made to isolate or demonstrate the aetiological organisms in 13 studies. The methods used in these studies for identification of bacteria were a blood culture, sputum examination or urinary antigen detection. For this review, results of a throat swab for bacterial isolation were ignored. Bacterial pathogens could be identified in blood cultures or serology/sputum in 513 (10.8%) out of 4742 patients tested. Out of the bacterial pathogens identified, 212 (41%) patients had S. pneumoniae, 129 (25%) had H. influenzae, 58 (11%) had Staphylococcus aureus (S. aureus) and 114 (22%) had other pathogens including the gram negative bacilli M. catarrhalis and Staphylococcus albus (S. albus) ( Table 2).
Information regarding the sensitivity pattern of bacterial isolates was available in four studies (Asghar 2008; Bansal 2006; Mulholland 1995; Roord 1996). This information was only available for the antibiotics studied and sensitivity was not tested in all the isolates. In the study by Asghar et al, out of a total of 22 S. pneumoniae isolates, 13/14 were sensitive to chloramphenicol, 12/17 to gentamycin, 15/16 to ampicillin and 12/12 to third generation cephalosporins (Asghar 2008). Out of a total of eight isolates of H. influenzae, 6/7 were sensitive to chloramphenicol, 12/17 to gentamicin, 15/16 to ampicillin and 6/6 to third generation cephalosporins. Out of a total of 47 isolates of Staphylococcus aureus 19/37 were sensitive to chloramphenicol, 29/45 to gentamycin, 15/16 to ampicillin and 6/6 to third generation cephalosporins. In the study by Bansal (Bansal 2006), all the three isolates of S. pneumoniae were sensitive to penicillin, amoxycillin, erythromycin and gentamycin. However, out of two isolates of H. influenzae, one was sensitive and the other isolate was resistant to penicillin, amoxycillin, erythromycin and gentamycin. The one that was resistant was sensitive to ciprofloxacin, cefotaxime and chloramphenicol. In the study by Mulholland (Mulholland 1995), all 10 isolates of S. pneumoniae were susceptible to co-trimoxazole and nine of these were also susceptible to chloramphenicol. All three Salmonella spp. isolates were susceptible to co-trimoxazole and chloramphenicol. A single isolate of H. influenzae was resistant to co-trimoxazole. In the study by Roord (Roord 1996), all 20 isolates were sensitive to azithromycin while three organisms were resistant to erythromycin.
Nasopharyngeal aspirates were tested for respiratory syncytial virus (RSV) in four studies (Atkinson 2007; Addo-Yobo 2004; Mulholland 1995; Wubbel 1999) involving 1916 children. RSV was identified in 403 children (21%). Identification of atypical organisms was attempted in five studies (Block 1995; Bradley 2007; Harris 1998; Kogan 2003; Wubbel 1999). Out of the 1594 patients tested for M. pneumoniae, 381 (24%) were tested positive. In patients aged under five years 141 out of 659 (21%) tested positive for mycoplasma. Tests for Chlamydia spp. were positive in 158 out of 1534 (10%) patients. In children under five years, there were positive test results for Chlamydia spp. in 45 out of 658 (7%) patients.
We attempted to compare various antibiotics (A and C) when comparisons of antibiotics A and B were available and B and C were available. We utilised this process to compare co-trimoxazole with co-amoxyclavulanic acid (Analysis 26), amoxycillin with cefpodoxime (Analysis 27) and amoxycillin with chloramphenicol (Analysis 28). Baseline data for age and sex were not comparable in the first two comparisons and therefore no valid comparison could be carried out. In the comparison of amoxycillin with chloramphenicol (CATCHUP 2002; Mulholland 1995; Straus 1998) sex distribution was not comparable although age distribution was. Cure rates were better in the amoxycillin group compared to the chloramphenicol group (OR 4.26; 95% CI 2.57 to 7.08) and failure rates were lower in the amoxycillin group (OR 0.64; 95% CI 0.41 to 1.00).
The aim of this review was to establish the most effective antibiotics for first-line empirical treatment of community-acquired pneumonia (CAP). A limited number of randomised controlled trials (RCTs) fulfilled the inclusion criteria. Most of the antibiotic comparisons were available in single studies only.
Summary of main results
Studies comparing ambulatory treatment of pneumonia suggest that co-amoxyclavulanic acid was better than amoxycillin. Amoxycillin and co-trimoxazole were associated with similar failure rates. Resolution of radiologic pneumonia was better with clarithromycin as compared to erythromycin and side effects were fewer with azithromycin as compared to co-amoxyclavulanic acid. For children with severe pneumonia treatment with oral amoxycillin was similar to that of injectable ampicillin or penicillin. Death rates were higher in children getting chloramphenicol as compared to those getting penicillin/ampicillin plus gentamycin.
For severe/very severe pneumonia, penicillin/ampicillin plus gentamycin was associated with lower re-admission rates as compared to chloramphenicol.
For very severe pneumonia failure rates were significantly higher as compared to ampicillin and gentamycin
The rest of the comparisons for ambulatory treatment involved azithromycin with erythromycin, clarithromycin, clarithromycin with erythromycin, amoxycillin with procaine penicillin, co-trimoxazole with single dose procaine penicillin followed by oral ampicillin, and cefpodoxime with co-amoxyclavulanic acid and there were no statistically significant differences in these comparisons.
Comparison for severe and very severe pneumonia involved chloramphenicol plus ampicillin with penicillin, amoxycillin with cefuroxime, amoxycillin with clarithromycin, penicillin and gentamycin with co-amoxyclavulanic acid, levofloxacin with ceftriaxone or co-amoxyclavulanic acid, cefuroxime with clarithromycin and chloramphenicol with co-trimoxazole and were comparable.
Overall completeness and applicability of evidence
Treatment of pneumonia depends on the age of the child, the severity of illness, the likely aetiological agents and their resistance pattern. The aetiological agents vary with age and possibly geographic location. Most of the studies included in this review were from under-developed countries with age groups below five years and identification of aetiological agents was limited to a few studies. The burden of pneumonia is significant in infants from developing countries. Attempts to isolate aetiological agents may not be cost-effective and therefore empirical treatment of pneumonia is justified. The results of this review may therefore be more applicable to the management of pneumonia in developing countries. However, data comparing two different antibiotics may also be useful in guiding antibiotic therapy in industrialised countries.
The World Health Organization (WHO) recommends treatment of non-severe pneumonia with co-trimoxazole as a first-line empirical antimicrobial treatment in countries with an infant mortality higher than 40 per 1000 live births (WHO 1991). Concerns about increasing resistance of common pathogens (S. Pneumoniae and H. Influenzae) to co-trimoxazole have been raised and amoxycillin has been suggested as an alternative. This review suggests that amoxycillin and co-trimoxazole are associated with similar failure rates. It can be concluded that there are insufficient data to show superiority of amoxycillin to co-trimoxazole. It should be noted that amoxycillin is more expensive than co-trimoxazole for five days of treatment for a child weighing between 5 and 10 kg (in India US $0.6 versus $0.3). Two recent studies (Agarwal 2004; MASCOT Group 2002) reported similar cure rates with amoxycillin given for three or five days. The cost of amoxycillin would be reduced to some extent if the treatment duration of amoxycillin was lowered to three days. Most studies comparing co-trimoxazole and amoxycillin used clinical case definition of pneumonia (rapid respiration). Respiratory symptoms and rapid respiratory rates in children may be due to bacterial pneumonia, viral infection associated wheeze, asthma etc. In a study from Pakistan chest radiographs were normal in 82% of children diagnosed with non-severe pneumonia using the WHO case definition (Hazir 2006). The majority of such children, except those with bacterial pneumonia, may not require antimicrobial agents and are likely to recover over three to seven days with supportive care. Giving them co-trimoxazole or amoxycillin or any other antibiotics may not alter their outcome. Therefore, it is important to have well-designed clinical trials in children with true pneumonia (radiologically confirmed/direct or indirect evidence of bacterial pneumonia).
Alternative antibiotics for community-acquired pneumonia include macrolides, co-amoxyclavulanic acid, cefpodoxime, procaine penicillin and benzathine penicillin. Comparison of various macrolides shows similar efficacy, with the exception of more radiological clearance with clarithromycin without any clinical implications. Macrolides may acquire resistance very fast if used indiscriminately (Inoue 2006). Amoxycillin was comparable with macrolides (azithromycin and clarithromycin), procaine penicillin and cefuroxime. Amoxycillin may therefore be preferable over these drugs. Co-amoxyclavulanic acid has been shown to give better results than amoxycillin and comparable results to azithromycin and cefpodoxime. The results are based on single study for each. This drug is relatively more expensive and may be used as a second-line drug. Cefpodoxime was comparable with co-amoxyclavulanic acid in a single study and may be an alternative second-line drug where co-amoxyclavulanic acid cannot be administered. Injectable penicillins (procaine penicillin or benzathine penicillin) are of limited use in non-severe pneumonia.
The WHO recommends admission to hospital and treatment with penicillin for severe pneumonia and chloramphenicol for very severe pneumonia (WHO 1999). In this review it emerged clearly that children with severe pneumonia without hypoxia who are feeding well can be treated with oral amoxycillin in hospital. The mortality rates were higher in children receiving injectable antibiotics. Quality assessment of three trials comparing oral with injectable medications reveals adequate allocation concealment but all were unblinded. There is no explanation for the increased death rates in those who received injectable antibiotics, as they were treated with either ampicillin/penicillin or amoxycillin. After excluding one study (Addo-Yobo 2004) that reported seven deaths in children receiving injections, the difference in death rate becomes non-significant. In view of the similar antimicrobial spectrum of all these drugs (ampicillin/amoxycillin/penicillin) and the possible benefit of better bioavailability with parenteral administration of antibiotics, a better outcome could be expected with use of injectable antibiotics for the treatment of children with severe pneumonia. The studies were carried out in a controlled environment (either in hospitals or with careful follow up) and so it is desirable to wait for more studies in community settings for ambulatory treatment of severe pneumonia with oral antibiotics. In children with severe or very severe pneumonia, it was evident that chloramphenicol was inferior to the combination of penicillin/ampicillin plus gentamycin. There is therefore a need to change the WHO guidelines. Alternative antibiotics for hospitalised children with severe and very severe pneumonia include ceftrioxone, levofloxacin, co-amoxyclavulanic acid and cefuroxime. However, comparisons were based on single studies and these drugs are relatively more expensive. They could therefore be considered for second-line therapy if a patient fails to respond to penicillin/ampicillin plus gentamycin.
Cure and failure rates of CAP depend not only on the choice of antibiotics but also on the aetiology of the pneumonia, the age of the patient, the sensitivity pattern of the bacterial pathogen, the severity of disease and any antibiotic usage in the recent past. While information on resistance patterns was not included in the studies evaluated in the review, this is likely to be of major importance in the future, in terms of both clinical practice and research.
In the management of CAP, isolation of bacterial pathogens in order to make a decision about the choice of antibiotics is not feasible in most circumstances. Even if bacterial pathogens are isolated, the child will need to be treated with empirical antibiotics until the result of the culture is available. In this review identification of bacterial pathogens was attempted in 13 studies (Asghar 2008; Bansal 2006; Block 1995; Bradley 2007; Camargos 1997; Duke 2002; Harris 1998; Klein 1995; Kogan 2003; Mulholland 1995; Roord 1996; Straus 1998; Wubbel 1999). Bacterial pathogens could be isolated in only 11% of the study participants. S. pneumoniae and H. influenzae constituted 67% of all the bacterial isolates. Therefore, empirical antibiotic therapy for CAP should be effective against these two pathogens.
Respiratory syncytial virus could be isolated in 21% of patients, suggesting that a sizeable proportion of patients may have a viral aetiology of CAP. These patients may not need antibiotics. A child with viral pneumonia can be identified from rapid diagnostic tests such as nasopharyngeal aspirates (Maitreyi 2000) and can avoid administration of antibiotics. However, the possibility of mixed infection (bacterial agents with viruses) has been observed in 10% to 40% of cases (Kabra 2003). At present, it is policy to treat all children with pneumonia with antibiotics due to a lack of tests that can reliably rule out bacterial pneumonia.
The aetiology of pneumonia depends on the age of the patient. In the present review the majority of enrolled subjects were below five years of age and separate data according to age were not available for primary and secondary outcomes in the studies that also enrolled older children. Therefore recommendations based on the evidence are more applicable to children below five years of age.
Another important issue is the aetiological role of atypical organisms (Chlamydia and Mycoplasma spp.) in CAP (Chaudhary 1998; Normann 1998; Pandey 2005). Five studies included in this review identified atypical organisms (Block 1995; Bradley 2007; Harris 1998; Kogan 2003; Wubbel 1999). Out of 1594 children tested for M. pneumoniae, 381 (24%) tested positive. The positivity for Mycoplasma in children under five years age was 21% (158/1534). Tests for Chlamydia spp. were positive in 158 out of the 1534 children (10%). In children under five years of age, positive tests for Chlamydia spp. occurred in 45 out of 658 (7%). The most effective antibiotics against atypical organisms are tetracycline and macrolides. In this review, the studies that attempted to identify atypical organisms showed equal cure rates between erythromycin and azithromycin. Two studies (Harris 1998; Wubbel 1999) comparing azithromycin with co-amoxyclavulanic acid in children under five years of age also showed equal cure and failure rates. In these studies the incidence of atypical organisms in children under five years of age was 15% and 11% for Mycoplasma spp. and Chlamydia, respectively. The cure rates in children receiving co-amoxyclavulanic acid were comparable to those receiving azithromycin. From this observation it can be inferred that either the diagnostic tests used for atypical organisms in these studies may not indicate invasive infections, or that the study was not adequately powered to detect small differences.
The sensitivity of the pathogens isolated in various studies was not available in all the studies. Wherever it was available it was not uniformly tested for common antibiotics.
Exposure to antibiotics in the recent past may adversely affect the outcome of bacterial pneumonia as the chances of infection with a resistant organism increases (Chenoweth 2000). In this review, information on past antibiotic use was available in seven studies (Addo-Yobo 2004; Asghar 2008; Atkinson 2007; Bradley 2007; Duke 2002; Hazir 2008; Straus 1998). The distribution of patients who had received antibiotics in the recent past was similar in the two treatment groups in all the studies. However, subgroup analysis was not available in these studies. In one study (Hazir 2008) antibiotic use in the last week was associated with increased failure rates on univariate analysis. In a study comparing co-trimoxazole and amoxicillin the number of patients who had received antibiotics in the recent past was higher in the amoxycillin group (34% compared with 25.6% in the co-trimoxazole group) (Straus 1998). The failure rates were higher in the co-trimoxazole group (19% compared with 16% in the amoxicillin group) (OR 1.33; 95% CI 1.05 to 1.67). The cure rates were better in the amoxycillin group compared to the co-trimoxazole group (81% in the co-trimoxazole group and 84% in the amoxycillin group) (OR 0.82; 95% CI 0.63 to 1.08) even though the proportion of children who had received antibiotics in the past week was higher in the amoxycillin group. This may indirectly indicate the superior efficacy of amoxycillin over co-trimoxazole in the treatment of pneumonia; evident even in those children who had recently received antibiotics.
Malnutrition may affect the treatment outcome of pneumonia. There was only one study in malnourished children (Mulholland 1995) which compared co-trimoxazole and chloramphenicol. The study did not show any significant difference in cure rates, failure rates or need for change in antibiotics.
There are limitations in reviewing antibiotic usage in CAP. Comparisons are often performed among groups of children for whom identification of aetiological agents is lacking. This means that if the distribution of viral cases is not uniform, the conclusions regarding the efficacy of antibiotics can be debatable. Several individual factors, such as malnutrition, can deeply modify the evolution of CAP and the response to antibiotic therapy. In the present review, only one study addressed this problem; it is highly probable that this issue can influence the correct evaluation of the data. No data regarding antibiotic resistance were reported in the majority of the studies. It is well known that in some cases the level of resistance to commonly used antibiotics can have a great influence on the response to therapy. The role of atypical bacteria in the determination of CAP in children living in low-income countries is not established, probably because the methods for identifying these pathogens are too complicated or too expensive, or both. These data are needed to more accurately define the best antibiotic therapy. The results may be more applicable for developing countries as most studies were done in these countries.
Quality of the evidence
Five out of 27 studies were double-blind studies and allocation concealment was adequate. Another 11 studies were unblinded but had adequate allocation concealment, classifying them as good quality studies. There was more than one study comparing co-trimoxazole with amoxycillin, oral amoxycillin with injectable penicillin/ampicillin and chloramphenicol with ampicillin/penicillin and studies were of good quality, suggesting the evidence for these comparisons is of high quality compared to other comparisons.
Potential biases in the review process
In this review we included one study (Awasthi 2008) of which one of the authors of present review (Kabra) was a co-author.
Agreements and disagreements with other studies or reviews
The important changes in this updated review in comparison to the previous version (Kabra 2006) include the following.
- The efficacy of co-trimoxazole and amoxycillin for ambulatory treatment of pneumonia was similar. This followed the inclusion of a study (Awasthi 2008) that compared three days of amoxycillin with five days of co-trimoxazole for CAP. In this study children receiving amoxycillin or co-trimoxazole were evaluated on day four and day six, respectively.
- Treatment of severe pneumonia with oral antibiotics in a controlled setting (hospitalised children or close follow up) is feasible.
- In very severe pneumonia penicillin/ampicillin plus gentamycin was superior to chloramphenicol.
A review comparing oral and injectable antibiotics in pneumonia suggested no difference in cure and failure rates in children getting oral or injectable antibiotics for the treatment of pneumonia in children (Rojas-Reyes 2006). In the present review we also found that oral and injectable antibiotics (amoxycillin) versus penicillin/ampicillin and co-trimoxazole versus procaine penicillin) for pneumonia are equally effective.
Implications for practice
For the treatment of ambulatory patients with CAP, amoxycillin is an alternative to co-trimoxazole. There are no apparent differences between azithromycin and erythromycin, azithromycin and co-amoxyclavulanic acid or cefpodoxime and co-amoxyclavulanic acid. There is limited data on other antibiotics: co-amoxyclavulanic acid and cefpodoxime may be alternative second-line drugs. Severe pneumonia without hypoxia and good oral acceptance can be managed in hospital/controlled settings with oral amoxycillin. Further community-based studies are required to determine the efficacy of oral antibiotics in the ambulatory management of severe pneumonia in the community. For children hospitalised with severe and very severe CAP, penicillin/ampicillin plus gentamycin is superior to chloramphenicol. The other alternative drugs for such patients are ceftrioxone, levofloxacin, co-amoxyclavulanic acid and cefuroxime. Until more studies are available these can be used as a second-line therapy.
More randomised controlled trials are required for a review of these antibiotics in order to make more accurate recommendations for their prescription.
Implications for research
There are many new antimicrobials available for the management of non-severe and severe CAP. There is a need for more studies, using similar methodologies and large numbers of patients, to compare amoxycillin with co-amoxyclavulanic acid, macrolides with amoxycillin and amoxycillin with oral cephalosporins. There is need for high quality trials to validate the efficacy of oral amoxycillin in the treatment of severe pneumonia in community settings.
We acknowledge all the help and infrastructure provided by the All India Institute of Medical Sciences, New Delhi, where all the authors serve as faculty. We acknowledge the help provided by Elizabeth Dooley, Managing Editor and Sarah Thorning, Trials Search Co-ordinator of the Cochrane Acute Respiratory Infections Group, for doing the EMBASE search and getting full text articles of studies. We also acknowledge the help provided by Dr Sunil Saharan in masking the study articles. We are very thankful to the referees Dr Roger Damoiseaux, Dr Yingfen Hsia, Dr Rajni Bhatia and Dr Max Bulsara for their input in improving the quality of this updated review.
Data and analyses
- Top of page
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Differences between protocol and review
- Index terms
Appendix 1. EMBASE search strategy
#1 explode 'pneumonia-' / all subheadings in DEM,DER,DRM,DRR
#2 (pneumonia in ti) or (pneumonia in ab)
#3 #1 or #2
#4 'antibiotic-agent' / all subheadings in DEM,DER,DRM,DRR
#5 (antibiotic* in ti) or (antibiotic* in ab)
#6 #4 or #5
#7 'child-' / all subheadings in DEM,DER,DRM,DRR
#8 (child in ti) or (child in ab)
#9 (children in ti) or (children in ab)
#10 'infant-' / all subheadings in DEM,DER,DRM,DRR
#11 (infant* in ti) or (infant* in ab)
#12 #7 or #8 or #9 or #10 or #11
#13 #3 and #6 and #12
#14 explode 'randomized-controlled-trial' / all subheadings
#15 explode 'controlled-study' / all subheadings
#16 explode 'single-blind-procedure' / all subheadings
#17 explode 'double-blind-procedure' / all subheadings
#18 explode 'crossover-procedure' / all subheadings
#19 explode 'phase-3-clinical-trial' / all subheadings
#20 (randomi?ed controlled trial in ti) or (randomi?ed controlled trial in ab)
#21 ((random* or placebo* or double-blind*)in ti) or ((random* or placebo* or double-blind*)in ab)
#22 (controlled clinical trial* in ti) or (controlled clinical trial* in ab)
#23 #14 or #15 or #16 or #17 or 318 or #19 or #290 or #21 or #22
#24 (nonhuman in der) not ((human in der) and (nonhuman in der))
#25 #23 not #24
#26 #13 and #25
Last assessed as up-to-date: 17 September 2009.
Protocol first published: Issue 3, 2004
Review first published: Issue 3, 2006
Contributions of authors
Dr Sushil K Kabra (SK) and Dr Rakesh Lodha (RL) jointly prepared and edited the review.
Dr RM Pandey (RP) contributed to the sections on data extraction, data analysis, quality assessment and statistical methods; in addition to editing the review.
Declarations of interest
One of the authors (Kabra) was co-author in one study (Awasthi 2008) included in the review.
Sources of support
- All India Institute of Medical Sciences, New Delhi, India.
- No sources of support supplied
Differences between protocol and review
In protocol we decided to include studies with an outcome in form of cure rate. However, there were a few studies that did not have cure rates. We therefore decided to include studies that gave either cure rates or treatment failure rates as one of the outcomes.
Medical Subject Headings (MeSH)
Anti-Bacterial Agents [*therapeutic use]; Community-Acquired Infections [drug therapy]; Drug Therapy, Combination [methods]; Pneumonia, Bacterial [*drug therapy]; Randomized Controlled Trials as Topic
MeSH check words
Adolescent; Child; Humans