The burden of Pneumocystis carinii pneumonia (PCP) has until recently been poorly described in African children infected with HIV-1. Autopsy studies indicated that 16–31% of HIV-1-infected children who died had PCP [1–3]. Although PCP was diagnosed in 45–48% of HIV-1-infected children investigated for pneumonia in developed countries before the institution of highly active antiretroviral treatment or cotrimoxazole prophylaxis, the burden of PCP in HIV-1-infected African children has only recently been quantified using the non-invasive techniques of nasopharyngeal aspirates (NPA) or induced sputum [4–7]. Compared with the high yield of 45% obtained from children in developed countries using bronchoalveolar lavage (BAL), the yield of P. carinii using induced sputum (10%) and NPA (17.2%) in two studies reported from Africa were lower [4,6,7]. Differences in the selection criteria for investigating for PCP and in the technique used for isolating P. carinii may account for the observed differences. Although BAL is considered to be the best method for diagnosing PCP, the usefulness of induced sputum or NPA has not been quantified. Because invasive tests such as BAL are unavailable in most developing countries, where the burden of HIV-1 infection is greatest, it is essential to evaluate alternative methods of diagnosing PCP, to be able to quantify the need for prophylaxis and to initiate appropriate antibiotic therapy when the burden of HIV-1 infection is high. The objectives of this study were to evaluate the burden of PCP in HIV-1-infected children who were hospitalized for severe pneumonia, to compare the reliability of induced sputum and NPA in isolating P. carinii, and to assess the usefulness of clinical and auxiliary laboratory tests in diagnosing PCP.
A prospective study was conducted at Chris Hani Baragwanath Hospital in Soweto, South Africa. This hospital provides secondary and tertiary level care to an estimated population of 1.2 million individuals, including 120 000 children under 5 years of age . The prevalence of HIV-1 infection among pregnant women attending antenatal clinics in 1999 was 23.4%. It was estimated that on the basis of a vertical transmission rate of 26%, at least 5% of the 22 000 children born at this hospital and its associated clinics in 1998–1999 would have been HIV-1 infected [9,10]. Neither mothers nor their children received antiretroviral treatment, and children are generally started on cotrimoxazole prophylaxis only after being hospitalized for an HIV-1-related illness.
Ethical clearance for the study was obtained from the Committee for Research on Human Subjects of the University of the Witwatersrand, and informed written consent for inclusion in the study and for HIV-1 testing was obtained from the parent of the child.
Children less than 2 years of age who were either known or clinically suspected of being HIV-1 infected and who were hospitalized for severe pneumonia were enrolled between February and June 1999. The criteria used for clinically suspecting possible HIV-1 infection included: (i) that the child had previously been diagnosed as being HIV-1 infected; (ii) the mother was known to be HIV-1 infected; or (iii) the child had three or more features of the Centers for Disease Control and Prevention (CDC) category A clinical criteria, or fulfilled the CDC clinical criteria of either categories B or C . Children were also required to have presented with a history of cough or shortness of breath, with either clinical evidence of severe respiratory distress as defined by the World Health Organization criteria , or a haemoglobin oxygen saturation of 90% or less on room air as measured by pulse oximetry. Children admitted with a clinical diagnosis of bronchiolitis were excluded from the study.
A physical examination was performed to quantify the severity of the pneumonia as well as to categorize clinically the HIV-1 severity based on the CDC clinical criteria for HIV-1-infected children . In addition, the parent/guardian or the child was interviewed regarding the clinical history as well as whether or not the child was receiving cotrimoxazole prophylaxis and whether or not the child was compliant. Children who were receiving cotrimoxazole prophylaxis were categorized either as being: (i) on adequate prophylaxis [i.e. had been receiving cotrimoxazole prophylaxis of adequate dosage (5 mg/kg of trimethroprim at least four times a week) for more than 2 weeks]; (ii) sub-optimal prophylaxis (i.e. had been started on cotrimoxazole less than 2 weeks before being hospitalized for the current illness or inadequate dose); and (iii) non-compliant (i.e. had stopped taking prophylaxis for more than one week or took prophylaxis less than four times a week).
The investigations performed in children enrolled into the study included: Testing for HIV-1 infection status whereby children were evaluated using two third-generation enzyme-linked immunosorbent assay (ELISA) tests (Axsym system, HIV 1/2, Abbot Diagnostics, Wiesbaden, Germany and the Wellcozyme HIV 1+2 EIA test, Murex Diagnostics, Kent, UK) and those younger than 15 months of age had their ELISA results confirmed using a qualitative DNA or a quantitative RNA-HIV-1 polymerase chain reaction (PCR) test (HIV-1 AMPLICOR kit, version 1.0, Roche Diagnostics, NJ, USA). Children less than 15 months of age with a reactive HIV-1 ELISA test but negative HIV-1 PCR were considered as being HIV-1 exposed but uninfected.
Additional blood tests that were performed included complete blood counts, lactate dehydrogenase, C-reactive protein, CD4 lymphocyte count and blood cultures using the BacT/Alert method.
Gastric washings were requested at the discretion of the attending physician to investigate for pulmonary tuberculosis using the Bactec method coupled with PCR identification for Mycobacterium tuberculosis using the IS 6110 as a probe as well as conventional culture techniques using Lowenstein–Jensen medium.
Additional investigations performed included chest radiographs that were interpreted by a paediatric radiologist as well as NPA, which were submitted for respiratory virus identification using direct immunofluorescent antibody testing and shell-vial culture techniques for respiratory syncytial virus, parainfluenza virus types 1–3, influenza virus types A and B and adenovirus. The technique for respiratory virus identification has been described previously .
Nasopharyngeal aspirate and induced sputum technique
NPA were performed using a modified feeding catheter (size FG8) that was attached to a 5 ml syringe containing 3–5 ml of normal saline. The feeding catheter was passed through either nostril into the nasopharynx and the saline was flushed through the syringe and immediately aspirated by applying suction with the syringe. The resultant aspirate was divided into two portions, one of which was added to viral transport medium and processed at the National Institute for Virology laboratory, South Africa, within 4 h for respiratory virus identification, and the remaining specimen was used for P. carinii identification at the South African Institute for Medical Research laboratory. NPA were performed before induced sputum sampling.
Induced sputum was collected by a registered physiotherapist who initially nebulized the child with 20 ml of 5% saline over 15–20 min, and the resulting expectorant was suctioned using a close-loop mucus trap that was attached to suction. Both methods were performed at least 4 h after a feed. The induced sputum was performed within 48 h of hospitalization. The children were observed for 20 min after the procedures for potential side-effects such as vomiting, epistaxis or deterioration in respiratory status.
P. carinii identification was performed using a direct monoclonal antibody immunofluorescent stain (MERIFLUOR Pneumocystis, Ohio, USA). A test was considered positive if more than two typical cysts exhibiting an apple-green fluorescence characteristic of P. carinii morphology was observed on 100 × magnification using dark-field microscopy.
Parental written consent was sought for limited tru-cut lung biopsies on those HIV-1-infected children who died during the course of their hospital stay. Samples were processed for histological evidence of PCP, which included Grocott–Gomori staining for P. carinii organisms. Furthermore, a diagnosis of cytomegalovirus (CMV) pneumonia was made by visualization of intranuclear viral inclusion bodies consistent with CMV infection and by immunohistochemistry for CMV. A diagnosis of interstitial pneumonitis was made in the presence of the following histological features: a thickening of the alveolar septae by a mononuclear inflammatory cell infiltrate including lymphocytes, plasma cells and histiocytes with spindle cell proliferation as well as the proliferation of type II pneumocytes.
The reason for deciding on needle biopsies rather than full biopsies was related to the lack of cultural acceptability of full biopsies in this community.
The data were analysed using Epi Info version 6.04c (Centers for Disease Control and Prevention, Atlanta, USA). Continuous variables were expressed as means and standard deviations (s.d.) or medians and ranges for parametric and non-parametric results, respectively. Categorical variables were analysed using the Mantel–Haenszel chi squared test or the Fisher's exact test if the expected cell value was less than five. In addition, the sensitivity, specificity, positive and negative predictive values of induced sputum and NPA were calculated using lung histology findings as a reference standard.
A total of 121 children were enrolled, of whom 119 were included in the analysis. The excluded children included one child who was older than 2 years of age and another who died before being investigated. HIV-1 infection was confirmed in 105 children whereas the remaining 14 children had a reactive ELISA test but a negative HIV-1 PCR test. Of the 105 HIV-1-infected children, 40 (38.1%) children had previously been diagnosed as being HIV-1 infected, two (1.9%) children were known to have been born to HIV-1-infected mothers and 73 (69.5%) children were included on the basis of clinical suspicion as per CDC clinical criteria A–C.
Pneumocystis carinii pneumonia in HIV-1-infected children
Induced sputum was performed in all 105 children whereas NPA were carried out in 101 (96.2%) of these children. P. carinii cysts were identified in 51 out of 105 (48.6%) children, either from induced sputum (37/105, 35.2%) or NPA (26/10, 25.7%). Of the children in whom PCP was diagnosed, P. carinii cysts were identified from both induced sputum and NPA in 12 out of the 48 (25%) children who had both procedures performed. Demographic information and clinical data on the HIV-1-infected children are shown in Table 1. Of the 40 children who had a history of being on cotrimoxazole prophylaxis, 31 (77.5%) had been on adequate prophylaxis, seven (17.5%) were on prophylaxis for less than 14 days and two (5%) children had defaulted prophylaxis for 8 and 14 days, respectively. Of the children who were receiving cotrimoxazole prophylaxis, PCP occurred in 13 children who were considered to be receiving adequate prophylaxis and in one child who had been on prophylaxis for less than 2 weeks’ duration. Overall, P. carinii was more likely to be isolated in children who had never received cotrimoxazole prophylaxis (P = 0.02). Among children in whom the use of cotrimoxazole prophylaxis was reported, there was no difference in the length of time that prophylaxis had been given between those children in whom P. carinii had and had not been isolated [median duration (range) 7.2 (0.4–67.5) weeks versus 7.5 (0.3–74.5) weeks, P = 0.37, respectively].
Furthermore, there were no other distinguishing features either on history or clinical examination that differentiated children with PCP from those without PCP (Table 2). Similarly, there were no laboratory markers including lactate dehydrogenase, leukocyte count, radiological findings and percentage CD4 lymphocyte count that distinguished between children with and without PCP (Table 2). A concurrent respiratory viral infection or bacteraemia was observed in nine out of 51 (17.6%) of the children with PCP (Table 2).
Adverse events observed as a result of induced sputum sampling included two out of 105 (1.9%) children with self-limiting epistaxis and three out of 105 (2.9%) children who experienced an episode of vomiting after sampling. The transient drop in oxygen saturation, as measured by a drop in oxygen saturation to less than 90% off oxygen supplementation or a clinical impression of cyanosis, was observed in 94 out of 105 (89.5%) of the children during suctioning associated with sampling by the induced sputum method, and lasted for less than 1 min after oxygen supplementation was restarted in all children.
Mortality and post-mortem lung biopsy findings in HIV-1-infected children
Mortality was high (27.6%) in children with and without PCP (Table 2). Of the 29 children who died, consent for post-mortem lung biopsies was obtained in 18 children, of whom eight (44.4%) children had histological evidence of PCP (Table 3). Interestingly, none out of eight (0%) and three out of 10 (30%) children with and without histological evidence of PCP, respectively, were on cotrimoxazole prophylaxis (P = 0.21) at the time of their hospitalization (Table 3). Furthermore, one or more pathogens that may have contributed to death was found in 14 out of 18 (77.8%) of the children in whom post-mortem lung biopsies were performed, whereas all of the other four children in whom no pathogen was identified had non-specific interstitial pneumonitis (Table 3). Overall, eight out of 18 (44.4%) children had evidence of CMV pneumonia on post-mortem lung histology. Seventy-five per cent (6/8) of the children with CMV pneumonia also had evidence of PCP, diagnosed either by NPA or induced sputum, including four who also had histological evidence of PCP (Table 3).
The sensitivity, specificity, positive predictive value and negative predictive value of induced sputum or NPA were compared with the post-mortem histological diagnosis of PCP (Fig. 1). Although there were no statistically significant differences in the sensitivity or specificity of induced sputum versus NPA or versus the results of both methods combined, the sensitivity was best when the results of both induced sputum and NPA were combined (Fig. 1).
Pneumocystis carinii pneumonia in HIV-1-exposed uninfected children
Fourteen children (11.8%) were identified as being HIV-1 exposed but uninfected on the basis of a non-reactive HIV-1 PCR test in the presence of two reactive HIV-1 ELISA tests. Seven of these children (50%) were diagnosed as having PCP, and the clinical and laboratory parameters of these seven children are detailed in Table 4. Malnutrition was diagnosed in five out of seven (71.4%) of these children. Furthermore, concurrent Pseudomonas aeruginosa, respiratory syncytial virus (RSV) or Mycobacterium tuberculosis pneumonia was diagnosed in three other children (Table 4). The organisms isolated in the remaining seven children without PCP included one isolate of Streptococcus pneumoniae from blood and two isolates of RSV from NPA. The one death among these 14 children occurred in a child without PCP.
Differences in the selection criteria as well as in the techniques used in obtaining samples when investigating for P. carinii may explain the higher burden of PCP (48.5%) observed in hospitalized children at our centre compared with the 10–17.2% yield recently reported from two other African studies [6,7]. There is conflicting data regarding the value of NPA in diagnosing PCP, with at least one study showing no value [6,7]. The technique used in the latter study, however, differed from that used in this study, and involved suctioning of the nasopharynx with a sterile suction catheter attached to a mucus extractor after the installation of four drops of sterile saline into the nostrils , whereas the technique used in the study from Malawi was not described . Furthermore, the selection of children in the two other African studies was based on WHO criteria for severity, and did not use pulse oximetry as an adjunct to screen children. Despite the limitation of only having obtained limited post-mortem lung biopsies in 18 out of the 29 children who died, the correlation between the non-invasive technique results and the histology findings support the usefulness of induced sputum and NPA in diagnosing PCP in children between 2 months and 2 years of age. The lower positive and negative predictive values of NPA may suggest a false-positive result, or alternately it might be that the affected lung segment was not sampled because of the limited lung biopsy sampling technique. Consequently, the sensitivity of NPA may be understated by our analysis. The sensitivity of induced sputum in diagnosing PCP observed in this study was similar to that reported for adults (65 versus 52–56%) .
A previous report from the study site evaluating the burden of severe pneumonia (based on clinical and laboratory criteria) estimated that approximately 20% of HIV-1-infected children hospitalized with severe pneumonia had PCP . The results from the present study indicate that not only was the burden of PCP underestimated, but that both clinical and laboratory parameters are unhelpful in distinguishing PCP from other causes of severe pneumonia. The large number of children with PCP, in whom concurrent bacterial and viral pathogens were isolated, may partly explain the lack of distinguishing markers between children with and without PCP when hospitalized for severe pneumonia. Although P. carinii was isolated less frequently in children receiving cotrimoxazole prophylaxis, at least 27.5% of children in whom PCP was diagnosed were receiving prophylaxis at the time of their presentation. This suggests that either the affected children were non-compliant, despite the history obtained from the escort, or alternatively that cotrimoxazole prophylaxis was ineffective in preventing PCP in these children who may have been severely immunocompromised. Importantly, however, none of the children who died that had histological evidence of PCP were on cotrimoxazole prophylaxis. The emergence of sulpha-resistant P. carinii after exposure to sulpha drugs, as has been described elsewhere , possibly needs to be evaluated in this setting.
The large percentage of children (44.4%) in this study in whom CMV pneumonia was diagnosed on post-mortem lung biopsies has previously been described in HIV-1-infected African children in Abidjan (also 44%) . The use of corticosteroids has been associated with an improved outcome in children with PCP . However, whether the empirical treatment of children suspected of being HIV-1 infected hospitalized for severe pneumonia with high-dose steroids (prednisone 2 mg/kg per day), as is practised at this centre, may partly contribute to the development or progression of CMV pneumonia needs further evaluation.
Furthermore, the high burden of PCP in HIV-1-exposed but uninfected children is intriguing. Although this may be related to a compromised immune system caused by the underlying malnutrition observed in most of these children who were diagnosed with PCP , the possible high-density shedding of P. carinii by their mothers, who may not have been on cotrimoxazole prophylaxis or who were not receiving antiretroviral treatment, may have contributed to this. Additional factors that may have predisposed these children to developing PCP would include an immature immune system in the affected neonate and other undiagnosed immune deficiencies. Although the causes for the high mortality rate observed in HIV-1-exposed uninfected children in Africa may be multi-factorial, the observation of PCP in this category of children also warrants further investigation.
We believe that previous reports in African HIV-1-infected children not receiving antiretroviral treatment may have underestimated the burden of PCP in those who were hospitalized for pneumonia. Furthermore, we have confirmed that induced sputum together with NPA are useful non-invasive techniques for diagnosing PCP. This should serve as a useful tool in evaluating the effectiveness of cotrimoxazole prophylaxis, which needs to be established urgently in sub-Saharan Africa, with its large burden of HIV-1-infected children to whom antiretroviral treatment is currently unavailable.
Induced sputum and NPA were performed by Edna Dahan, chest radiographs were interpreted by Dr Elise Cumin, CD4 lymphocyte counts were processed by Dr Gayle Sherman and Ms Lesley Scott, and lung histology was analysed by Dr Martin Hayle. Dr Haroon Saloojee is acknowledged for his critical review of the manuscript. D.R. was involved in the protocol design, the recruitment of the subjects and write-up of the report. S.A.M. was involved in the protocol-design, statistical analysis and write-up of the report.
1. Lucas SB, Peacock CS, Hounnou A. et al
. Disease in children infected with HIV in Abidjan, Cote d'Ivoire. BMJ 1996, 312: 335–338.
2. Ikeogu MO, Wolf B, Mathe S. Pulmonary manifestations in HIV seropositivity and malnutrition in Zimbabwe. Arch Dis Child 1997, 76: 124–128.
3. Friedland IR, Klugman KP, Karstaedt AS, Patel J, McIntyre JA, Allwood CW. AIDS – the Baragwanath experience. Part 1: Epidemiology of HIV infection at Baragwanath Hospital; 1988–1990.
South Afr Med J 1991, 82: 86–90.
4. Gibb DM, Davison CF, Holland FJ, Walters S, Novelli V, Mok J. Pneumocystis carinii
pneumonia in vertically acquired HIV infection in the British Isles. Arch Dis Child 1994, 70: 241–244.
5. Foot ABM, Caul EO, Roome AP, Oakhill A, Catterall JR. An assessment of sputum induction as an aid to diagnosis of respiratory infections in the immunocompromised child. J Infect 1992, 24: 49–54.
6. Graham SM, Mtitimila EI, Kamanga HS, Walsh AL, Hart CA, Molyneux ME. Clinical presentation and outcome of Pneumocystis carinii
pneumonia in Malawian children. Lancet 2000, 355: 369–373.
7. Zar HJ, Dechaboon A, Hanslo D, Apolles P, Magnus KG, Hussey G. Pneumocystis carinii
pneumonia in South African children infected with human immunodeficiency virus. Pediatr Infect Dis J 2000, 19: 603–607.
8. Daponte BO. Results from the 1995 Soweto population survey: estimates of the population of towns in Soweto
. Report to the Greater Johannesburg Metropolitan Council
. Pretoria, South Africa: Human Rights Institute of South Africa; July 1995.
9. Health Systems Research and Epidemiology. Summary report: National HIV sero-prevalence survey of women attending public antenatal clinics in South Africa.
Pretoria, South Africa: Department of Health; April 2000.
10. Gray GE, McIntyre JA, Lyons SF. Effect of breastfeeding on the vertical transmission of HIV-1 in Soweto, South Africa.
In:XIth International Conference on AIDS
. Vancouver, July 12–17 1997 [Abstract Th.C.413].
11. World Health Organization. Oxygen therapy for acute respiratory infections in young children in developing countries.
Geneva: WHO (WHO/ari/93.28); 1993.
12. Centers for Disease Control and Prevention. 1994 Revised classification system for human immunodeficiency virus infection in children less than 13 years of age. MMWR 1994, 43: 1–19.
13. Madhi SA, Schoub B, Simmank K, Blackburn N, Klugman K. Increased burden of respiratory viral associated severe lower respiratory tract infections in children infected with human immunodeficiency virus type 1. J Pediatr 2000, 137: 78–84.
14. Metersky ML, Aslenzadeh J, Stelmach P. A comparison of induced and expectorated sputum for the diagnosis of Pneumocystis carinii
pneumonia. Chest 1998, 113: 1555–1559.
15. Madhi SA, Petersen K, Madhi A, Khoosal M, Klugman KP. Increased disease burden and antibiotic resistance of bacteria causing severe community-acquired lower respiratory tract infections in human immunodeficiency virus type 1-infected children. Clin Infect Dis 2000, 31: 170–176.
16. Helweg-Larsen J, Benfield TL, Eugen-Olsen J, Lundgren JD, Lundgren B. Effects of mutations in Pneumocystis carinii
dihydropteroate synthase gene on outcome of AIDS-associated P. carinii
pneumonia. Lancet 1999, 354: 1318–1319.
17. Bye MR, Cairns-Bazarian AM, Ewig JM. Markedly reduced mortality associated with corticosteroid therapy of Pneumocystis carinii
pneumonia in children with acquired immunodeficiency syndrome. Arch Pediatr Adolesc Med 1994, 148: 638–641.
18. Hughes WT, Price RA, Sisko F. et al
. Protein-calorie malnutrition. A host determinant forPneumocystis cariniiinfection.
Am J Dis Child 1974, 128: 44–52.