Treatment failure in severe pneumonia highlights need to revise WHO guidelines for acute respiratory infections in South African children

This article is more than 17 years old. Click here for more recent articles on this topic

Pre-HIV World Health Organization (WHO) guidelines for managing acute respiratory infections in children younger than one year old need to be revised, according to the findings of a prospective study published in the April 28th edition of The Lancet. The study found that polymicrobial disease was a significant cause for treatment failure in children born to HIV-infected mothers.

Acute respiratory infections are the leading cause of childhood mortality worldwide. The emergence of HIV/AIDS has exacerbated the incidence of severe childhood pneumonia. In South Africa, HIV-related pneumonia is a leading cause of hospital admissions in 11-45 % of children. The management of this disease is constrained by pre-HIV WHO treatment guidelines.

These guidelines have been adapted so that all children with severe pneumonia receive high dose benzylpenicillin and gentamicin and high-dose trimethoprim sulfamethoxazole for pneumocystis. However, there is a paucity of field studies on the usefulness of the guidelines. In order to address this gap, a team of British and South African investigators carried out a prospective study of children hospitalised with severe pneumonia and receiving standard antimicrobial therapy.


treatment failure

Inability of a medical therapy to achieve the desired results. 


Any lung infection that causes inflammation. The infecting organism may be bacteria (such as Streptococcus pneumoniae), a virus (such as influenza), a fungus (such as Pneumocystis pneumonia or PCP) or something else. The disease is sometimes characterised by where the infection was acquired: in the community, in hospital or in a nursing home.


In a bacteria culture test, a sample of urine, blood, sputum or another substance is taken from the patient. The cells are put in a specific environment in a laboratory to encourage cell growth and to allow the specific type of bacteria to be identified. Culture can be used to identify the TB bacteria, but is a more complex, slow and expensive method than others.

enzyme-linked immunosorbent assay (ELISA)

A diagnostic test in which a signal produced by an enzymatic reaction is used to detect and quantify the amount of a specific substance in a solution. Can be used to detect antibodies to HIV, p24 antigen or other substances.

prospective study

A type of longitudinal study in which people join the study and information is then collected on them for several weeks, months or years. 

The study took place in Durban, KwaZulu-Natal, when the provincial antenatal HIV prevalence was 36.5 % and antiretrovirals were not available. Haemophilus influenzae type B vaccine was introduced in July 1999.

The eligible study population were children aged 1-59 months who were admitted with WHO-defined severe pneumonia with or without severe disease as diagnosed using non-bronchoscopic bronchoalveolar lavage (NB-BAL) between January 2001 and December 2002, and whose parents consented.

A detailed history was taken and a physical examination carried out. Laboratory investigations included a full blood count, blood culture, chest radiograph, nasopharyngeal aspirate for viral culture, urine for antimicrobial activity, induced sputum for pneumocytis and tuberculosis, three gastric washes for tuberculosis, and an HIV test. Pleural effusion was tapped and sent for microscopy and culture.

Infants under one year were classified as HIV-infected if they were ELISA positive and RNA positive, HIV-exposed, uninfected if they were HIV ELISA positive but RNA and DNA negative, and HIV-uninfected if they were ELISA negative. Children aged one year and over were classified into those infected and those uninfected.

Children received the study drug for a minimum of 48 hours and changed to oral treatment when no longer hypoxic. Dexamethasone was given every six hours intravenously in children with proven pneumocystis. Oxygen saturation was maintained above 92 %. A study doctor assessed the children within twelve hours of admission and twice daily until discharged.

The primary end-point was treatment failure at 48 hours; secondary end-points were cumulative treatment failure at any time during hospital admission and death. Children failing treatment were either transferred to the paediatric intensive care unit for further investigations or had the antibiotics empirically changed. They also had a repeat blood culture, nasopharyngeal and oropharyngeal swabs, and a chest radiograph.

Laboratory analysis involved Gram-stain microscopy, direct immunoflourescence, and culture of pleural fluid, BAL fluid, lung aspirates, induced sputa, and gastric washings. The focus of laboratory investigations was acid-fast bacilli, pneumocystis, and viruses such as respiratory syncytial virus, influenza A or B virus, parainfluenza virus types 1-3, cytomegalovirus, and adenovirus.

Out of the 358 children enrolled, 356 were black and two were Indians; 254 (74%) had WHO-defined very severe disease. Two hundred and forty-two (68 %) were HIV-1-infected, 41 (11 %) were HIV-1-exposed, and 75 (21 %) were uninfected. In general, HIV-exposed uninfected infants were better off than HIV-infected infants.

Thirty-five percent of the children failed treatment by 48 hours and a further 29 (8%) failed subsequently. Several factors including aged younger than one year, HIV-1 status, maternal tuberculosis status, polymicrobial disease, and very severe disease were predictors of poor treatment outcome. This was confirmed by logistic regression. Inability to drink, HIV ELISA positive, and admission oxygen saturation also predicted 48-hour treatment failure.

Infants under one year had higher 48-hour (42 % versus 17 %; p < 0.0001) and cumulative (49 % versus 30 %; p = 0.001) treatment failure rates, and higher in-hospital mortality rates (19 % versus 6 %; p = 0.004) than children aged one year or older. For children one year and older, there was no relationship between treatment outcomes and HIV status.

Both HIV-infected and HIV-exposed, uninfected infants had higher 48-hour and cumulative treatment failure rates than HIV-uninfected infants. HIV-infected infants tended to have higher 48-hour and cumulative treatment failure rates and in-hospital mortality than HIV-exposed uninfected infants. HIV-exposed, uninfected infants had significantly higher 48-hour treatment failure rate but not higher in-hospital mortality than HIV-uninfected infants. HIV status had an impact only in children with very severe disease.

Forty percent of children had urine antimicrobial activity at admission, indicating recent antimicrobial use. There was discordance between positive urinary antibiotic activity and recall of antibiotic use.

Thirty five percent of the children had a family history of tuberculosis within the past two years; only maternal tuberculosis was associated with poor outcomes. Bacteraemic children one year or older had worse outcomes than non-bactaeremic children. However, bacteraemia had no effect in children younger than one year.

Fifty-three children had culture positive tuberculosis which had no association with HIV status, age, or mortality. S. pneumoniae, S. aureus, H.influenze and M. tuberculosis were isolated from admission pleural effusions. Ninety-three percent of children with treatment failure had an organism isolated from at least one site. A number of bacterial species and a virus were isolated from 185 specimens. The commonest organisms were cytomegalovirus, Pneumocystis jirovecii, and M. tuberculosis.

In summary, risk factors for failure to respond to antibiotics by 48 h were identified in a cohort of South African children with severe pneumonia. These findings have important policy implications.

The current WHO guidelines for treatment of severe pneumonia are obsolete and require urgent revision. The increased risk in HIV-exposed uninfected infants is an important public health challenge since prevention of mother to child transmission programs will increase this population. Finally, polymicrobial disease is a key predictor of treatment failure. This calls for the deployment of rapid low-cost diagnostics.


McNally LM et al. Effect of age, polymicrobial disease, and maternal HIV status on treatment response and cause of severe pneumonia in South African children: a prospective descriptive study. Lancet 369: 1440-1451, 2007.