Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Pneumocystis jirovecii Pneumonia in Tropical and Low and Middle Income Countries: A Systematic Review and Meta-Regression

  • David M. Lowe ,

    Affiliations Wellcome Trust Centre for Research in Clinical Tropical Medicine, Department of Medicine, Imperial College London, London, United Kingdom, Clinical Infectious Disease Research Initiative, Institute of Infectious Diseases and Molecular Medicine, University of Cape Town, Cape Town, South Africa

  • Molebogeng X. Rangaka,

    Affiliation Clinical Infectious Disease Research Initiative, Institute of Infectious Diseases and Molecular Medicine, University of Cape Town, Cape Town, South Africa

  • Fabiana Gordon,

    Affiliation Statistical Advisory Service, Imperial College London, London, United Kingdom

  • Chris D. James,

    Current address: Oxford Policy Management, Oxford, United Kingdom

    Affiliation World Health Organisation Western Pacific Regional Office, Manila, Philippines

  • Robert F. Miller

    Affiliations Clinical Research Department, Faculty of Infectious and Tropical Diseases, London School of Hygiene and Tropical Medicine, London, United Kingdom, Research Department of Infection and Population Health, Institute of Epidemiology and Healthcare, University College London, London, United Kingdom

Pneumocystis jirovecii Pneumonia in Tropical and Low and Middle Income Countries: A Systematic Review and Meta-Regression

  • David M. Lowe, 
  • Molebogeng X. Rangaka, 
  • Fabiana Gordon, 
  • Chris D. James, 
  • Robert F. Miller



Pneumocystis jirovecii pneumonia (PCP), the commonest opportunistic infection in HIV-infected patients in the developed world, is less commonly described in tropical and low and middle income countries (LMIC). We sought to investigate predictors of PCP in these settings.


Systematic review and meta-regression.


Meta-regression of predictors of PCP diagnosis (33 studies). Qualitative and quantitative assessment of recorded CD4 counts, receipt of prophylaxis and antiretrovirals, sensitivity and specificity of clinical signs and symptoms for PCP, co-infection with other pathogens, and case fatality (117 studies).


The most significant predictor of PCP was per capita Gross Domestic Product, which showed strong linear association with odds of PCP diagnosis (p<0.0001). This was not explained by study design or diagnostic quality. Geographical area, population age, study setting and year of study also contributed to risk of PCP. Co-infection was common (444 episodes/1425 PCP cases), frequently with virulent organisms. The predictive value of symptoms, signs or simple tests in LMIC settings for diagnosis of PCP was poor. Case fatality was >30%; treatment was largely appropriate. Prophylaxis appeared to reduce the risk for development of PCP, however 24% of children with PCP were receiving prophylaxis. CD4 counts at presentation with PCP were usually <200×103/ml.


There is a positive relationship between GDP and risk of PCP diagnosis. Although failure to diagnose infection in poorer countries may contribute to this, we also hypothesise that poverty exposes at-risk patients to a wide range of infections and that the relatively non-pathogenic P. jirovecii is therefore under-represented. As LMIC develop economically they eliminate the conditions underlying transmission of virulent infection: P. jirovecii, ubiquitous in all settings, then becomes a greater relative threat.


Pneumocystis jirovecii pneumonia (PCP) is one of the commonest opportunistic infection in Human Immunodeficiency Virus (HIV)-infected patients in the developed world (historically the leading AIDS-indicator disease in both the United States and Europe) [1], but has always been regarded as rare in low and middle income countries (LMIC) predominantly of tropical latitude [2], [3]. However, the organism is ubiquitous and seroprevalence data indicates worldwide exposure [4][7]. This suggests that epidemiological differences are not explained by geography, nor are they easily attributable to genetic differences in the organism or the host [8], [9]. Factors influencing prevalence in these settings have not been systematically evaluated.

Our primary objective was to establish major determinants of PCP in HIV-infected individuals in tropical and LMIC. The ‘gold’ diagnostic standard of cytological examination of respiratory samples may be less available in LMIC, affecting reported prevalence. Gross domestic product (GDP), a measure of population income level, influences prevalence of HIV and opportunistic infections including tuberculosis [10], [11], and an association with PCP may thus be discovered. Relationships with time, geographical area or study setting may reflect genuine epidemiology [12], but might also reflect varying awareness of the condition among clinicians [13].

Additionally, assessment of the accuracy of symptoms, signs, or laboratory tests in LMIC for diagnosis of PCP, rates of co-infection that may confound diagnosis [14], PCP case fatality, CD4 count at presentation, treatment strategies, and efficacy of prophylaxis or antiretroviral therapy (ART) in reducing risk of PCP was undertaken.


Search Strategy and Study Selection

Table 1 details the search strategy. Summaries of AIDS-defining illnesses (ADI) or hospitalisations in the HIV-infected population were included for primary analysis, since these were considered most representative of overall disease burden and their broad entry criteria carried minimal risk of bias. Presumptive diagnoses were included, consistent with practice in many centres [1], although quality of diagnosis was also assessed. All HIV-infected participants in the studies were included. DML extracted and RFM verified the data.

Quality Assessment

A modified Newcastle Ottawa Scale (NOS) was used to assess selection and comparability of patients (Table 2). Since most studies in the primary analysis did not control for specific features, comparability was assessed on the basis of identical recruitment of PCP and non-PCP cases. Quality of outcome ascertainment was assessed by a quantitative diagnostic scoring system (Table 2). Clinical diagnosis only scored ‘1’; diagnosis based solely on spontaneously expectorated sputum scored ‘2’. Values between 3 and 5 indicate that appropriate advanced diagnostics were available and used for some, but not all, patients. A score of ‘6’ was awarded if all patients were diagnosed with high accuracy. Polymerase Chain Reaction (PCR) methods for diagnosing PCP were only being validated towards the end of the time range of these studies. Correspondingly, accuracy was not established and no studies used PCR alone without either immunofluorescence or cytology for diagnosis: for these reasons, PCR is not included in the scoring system.

Bias and Confounding

Within-study bias mostly derived from entry criteria specifying pre-existing diseases or clinical characteristics. Studies with biased entry criteria were excluded from pooled analysis. Increased clinical diagnosis of PCP resulting from greater awareness among clinicians may have created bias, but we could not formally assess this. We planned to control for time and geographical area in an attempt to eliminate or identify this bias. Population characteristics may vary depending on the type of research institution and so we also aimed to control for this variable.

Analysis and Statistics

Primary analysis was performed using logistic regression. The dependent variable was diagnosis of PCP (present/absent). Categorical predictor variables were study population (adult, paediatric, combined, unspecified); geographical area (Africa, Middle East, Indian sub-continent, East/South-East Asia, Central/South America) and study setting (rural/primary care, hospital, tertiary/specialist/University hospital, nationwide). Quality assessments (selection score and diagnostic score) were treated as categorical predictors. Due to low numbers within some diagnostic score groups, values were combined into categories of ‘Low quality’ (score 1 or 2), ‘High quality’ (score 5 or 6) or ‘Not Recorded’. Per capita GDP in ‘constant US dollars’ (after controlling for inflation) was obtained from the World Bank’s world development indicators for the median year of the study [15]. Median year of study and GDP were treated as continuous covariates. Year was centred to its mean; GDP showed a skewed distribution and was centred to its median. When the effects of GDP and year were seen to be non-linear a quadratic term was added. Relative importance of model components was assessed via Chi-square improvement (based on -2×Log Likelihood) with iterative addition of predictors.

Case fatality was assessed using logistic regression with death/survival as the dependent variable and the same predictors (as above). Odds Ratios for the effect of prophylaxis were calculated using Mantel-Haenszel methodology. A forest plot was used to assess heterogeneity in prevalence in studies selected for primary analysis; heterogeneity in case fatality and prophylaxis data was assessed via I2 analysis. Statistics were performed using SPSS version 18.0; meta-analysis functions were performed with Review Manager Version 5.1 (Cochrane Collaboration 2011, Copenhagen).


Of 1708 citations identified, 134 full-text papers were reviewed. Ten contained no identifiable data and seven were subsets of data published elsewhere (see Figure 1 for a PRISMA [16] flowchart; see Checklist S1 for the PRISMA checklist). Table 3 summarises the remaining 117 studies. Papers reviewed but not specifically cited in the text are included as Supplementary References (References S1). Eleven studies were from high income countries and 106 were from LMIC. Seventy-eight studies (67%) were performed in University/tertiary/specialist centres, 25 (21%) were performed in other (or undefined) hospital settings, five (4%) in rural healthcare facilities, seven (6%) were nationwide/regional studies and two (2%) reported occupational health services at mines.

Forty-six studies declared their treatment for PCP (Figure 2): 39 used co-trimoxazole, of which 19 mentioned concomitant use of steroids; four used pentamidine, two used clindamycin plus primaquine or pyrimethamine and one used dapsone. There were few data on the effect of ART on PCP prevalence. Several papers documented that no patients were taking ART at the time of presentation [17][21]. Lian et al [22] demonstrated a decline in PCP incidence from 21/128 (16.4%) prior to availability of ART to 1/128 (0.78%) subsequently; Kouakoussui et al [23] reported two cases of PCP in their cohort before availability of ART and none after. However, Brazilian mortality data [24], reported that 81 of 403 (est.) deceased PCP patients had received ART.

Figure 2. Treatment strategies for PCP in reviewed studies.

Total strategies for this analysis = 46.

PCP Prevalence is more Dependent on per capita Gross Domestic Product than any Other Factor

Thirty-five studies detailed ADI or reasons for hospitalisation within an HIV-infected population, but two pre-specified ‘fever of unknown origin’ and were excluded from primary analysis due to recruitment bias. The remaining 33 studies represented 114,389 people and 21,853 PCP diagnoses. However, one Thai surveillance study [25] accounted for 101,945 people and 20,145 diagnoses. Results are reported including this study, since removing it from analysis did not significantly alter findings. The overall median year of study was 1996 and median GDP was $2141 per capita. The prevalence of PCP as an ADI/reason for hospitalisation ranged from 1.67% to 60.0%. Initial assessment of heterogeneity (Figure 3) suggested that summary measures were inappropriate and we proceeded to meta-regression.

Figure 3. Prevalence of PCP in studies included for primary analysis, indicating heterogeneity of data set.

Confidence intervals for point prevalences were determined by adjusted Wald method.

There was a positive linear relationship between median year of study and odds of PCP (adjusted odds ratio (aOR) per year = 1.05, 95% CI 1.03–1.07, p<0.001). Geographical area was a significant predictor, with higher odds for a diagnosis of PCP in Central/South America (aOR = 1.72, 95% CI 1.35–2.19, p<0.001) and the Middle East (aOR = 2.53, 95% CI 1.46–4.39, p = 0.001) versus East/South-East Asia. However, there were only two studies from the Middle East (total n = 63), and so we interpreted this result cautiously. Adult or combined-age studies yielded lower odds for PCP than paediatric studies (for adult studies aOR = 0.67, 95% CI 0.52–0.86, p = 0.002; for combined studies aOR = 0.70, 95% CI 0.56–0.87, p = 0.001). Nationwide surveillance demonstrated higher odds than hospital studies (secondary or tertiary care), with the lowest odds in rural/primary care centres, although only one rural study (n = 225) was included (versus nationwide studies: aOR for tertiary/University hospitals = 0.46, 95% CI 0.35–0.60, p<0.001; aOR for other hospitals = 0.47, 95% CI 0.33–0.68, p<0.001; aOR for rural/primary care studies = 0.28, 95% CI 0.16–0.50, p<0.001).

GDP was a highly significant linear predictor of a diagnosis of PCP (p<0.0001). An increase of $20,000 (the approximate range of GDPs included) increases the odds for a diagnosis of PCP 10.54-fold (95% CI 8.41–13.20). Chi-square improvement analysis revealed GDP to be the most important predictor of a diagnosis of PCP (order of importance: GDP>population age>geographical area>study setting>median year). Results from logistic regression are summarised in Table 4. Figure 4A provides a graphical representation of prevalence data compared with GDP for adult studies; tuberculosis (TB) prevalence in the same studies is depicted in Figure 4B and shows a decline with increasing GDP, as previously described [11]. In these data, the correlation between PCP prevalence and GDP was 0.78 (p<0.001) unweighted, or 0.94 (p<0.001) weighted by study size; the correlation between TB prevalence and GDP was -0.61 (p = 0.001) unweighted, or -0.57 (p<0.001) weighted by study size.

Figure 4. Relationship between prevalence of diagnoses and GDP.

A. Relationship between Pneumocystis jirovecii pneumonia (PCP) prevalence (%) and per capita Gross Domestic Product (constant US$) in adult AIDS-defining illness studies or reasons for hospitalisation in an HIV infected population; size of marker indicates size of study; a regression line is added for clarity. Note the logarithmic x-axis. B. Relationship between tuberculosis (TB) prevalence (%) and per capita Gross Domestic Product (constant US$) for the same studies.

Table 4. Results of primary analysis of predictors of PCP via logistic regression.

Since Africa was under-represented in these data we also analysed the most common African study type, adult clinical studies of sub-acute respiratory illness (note there is some heterogeneity in entry criteria). There were 2109 HIV-infected patients and 277 PCP cases from 18 studies. There was a linear relationship between GDP and odds of PCP (p = 0.021); an increase of $20,000 increased the odds for PCP 37.2-fold (95% CI 1.74–794.2). There were lower odds for PCP in rural settings than hospitals (aOR = 0.28, 95% CI 0.08–0.96, p = 0.042) and a quadratic relationship with time showing a fall in odds followed by a rise over the included years (Table 5).

Table 5. Results of analysis of predictors of PCP via logistic regression for 18 African adult clinical respiratory studies.

Quality Assessments do not Explain the Associations Observed

Applying our modified NOS to the primary analysis, we concluded that patient comparability was acceptable since all studies recruited PCP and non-PCP cases identically (Table 2). Each study was also assessed according to the first three Selection parameters and awarded a score from 0 to 3. Selection score was not a significant predictor of PCP diagnosis in logistic regression, and a strong relationship with GDP persisted when analysis was restricted to studies which met two or more criteria (24 studies).

Diagnostic score was a significant predictor of PCP diagnosis in logistic regression (p<0.001). Lowest odds were seen in the ‘Low quality’ group, but both ‘Low quality’ and ‘High quality’ studies showed lower odds than the ‘Not Recorded’ group. However, after including diagnostic score as a predictor the only differences in other results were that the relationship with time became quadratic, plateauing between 2000–2001 (p = 0.002), and that lower odds of PCP were seen in Indian studies compared with East/South-East Asia (Table 4). GDP remained the strongest predictor using Chi-square improvement analysis; an increase in GDP of $1,000 in this model increased the odds for diagnosis of PCP 1.10-fold (95% CI 1.09–1.12). Restricting analysis to studies with high diagnostic quality (7 studies with score 5 or 6) revealed similar results to the overall analysis: an increase in GDP of $1,000 increased the odds for diagnosis of PCP 1.14-fold (95% CI 1.09–1.20, p<0.001). The African clinical studies generally diagnosed PCP with a high degree of certainty, with a Diagnostic Score of ‘6’ for 15 studies, ‘5’ for two studies and ‘2’ for only one study. No correlations existed between quality scores and either GDP or median year, including quality scores as fixed factors and controlling for other predictors in a linear regression model. In particular, there was no significant relationship between GDP and diagnostic score.

Diagnostic Value of Clinical Symptoms, Signs or Simple Tests are Generally Poor for PCP in Low Income Settings

We evaluated individual symptoms, signs, diagnostic tests and aspects of patient history for their ability to predict the presence or absence of a PCP diagnosis. Seven adult studies [26][32] (total 701 patients, 177 with PCP) and five paediatric studies [21], [33][36] (total 550 patients, 192 with PCP) were included; parameters detailed in three or more papers are displayed in Figure 5. Pooled estimates are not provided due to high heterogeneity. Indicators of PCP with high sensitivity and negative predictive value, for example cough or dyspnoea, generally had poor specificity and positive predictive value. Certain indicators pointing against a diagnosis of PCP, for example haemoptysis, demonstrated high specificity for a subset without PCP, but poor overall sensitivity for ‘PCP-negatives’.

Figure 5. Sensitivity and specificity for Pneumocystis jirovecii pneumonia (PCP) of most commonly described signs, symptoms, clinical tests or aspects of history.

A. predictive of PCP in adults, B. predictive of absence of PCP in adults, C. predictive of PCP in children, D. predictive of absence of PCP in children. * Chest x-ray descriptions most commonly associated with PCP in adults: ‘interstitial shadowing’; ‘fine shadowing’; ‘minimally abnormal not TB’; ‘diffuse shadowing’. Chest x-ray descriptions most commonly associated with PCP in children: ‘alveolar consolidation’; ‘interstitial infiltration’; ‘diffuse bilateral alveolar shadowing’; ‘consolidation’. ** Chest x-ray descriptions least commonly associated with PCP in adults: ‘cavities’; ‘hilar lymph nodes’; ‘classical TB’. TP = True Positive; FP = False Positive; FN = False Negative; TN = True Negative.

Co-infection is Common with PCP in Tropical and LMIC Settings

Table 6 demonstrates documented co-infections among 1,425 PCP patients, derived from 49 studies. These represent infectious rather than patient episodes and frequently a single patient had multiple co-existing infections. Studies whose entry criteria depended on a prior infectious diagnosis other than HIV (eg tuberculosis) were excluded from this analysis. There was a bias towards respiratory pathogens, with many studies basing all analysis on sputum/bronchial washings or including only patients with respiratory symptoms. M.tuberculosis accounted for at least 103 of 444 episodes (23.2%). Cytomegalovirus (CMV) was found in 104 episodes (23.4%). Bacterial infection (generally with virulent organisms, where specified) was identified in 116 episodes (26.1%).

Case Fatality from PCP in Tropical and LMIC Settings is High, is not Improving with Time, but Negatively Correlates with GDP

Table 7 details case fatality rates among people with a PCP diagnosis in the included studies. Because entry criteria differed between geographical areas, it was not considered valid to compare case fatality rates between regions. There was considerable heterogeneity in these data (I2>90% for all datasets except adult mortality in Central/South America), making us interpret pooled estimates cautiously. Nevertheless, as an indicative figure the overall case fatality from all studies in LMIC was 226/733 (30.9%, 95% CI 27.6%–34.3%). Limiting analysis to studies with diagnostic score 5 or 6 (22 studies) revealed a pooled case fatality estimate of 31.4%, and limiting to confirmed diagnoses (Score 6; 16 studies) revealed a pooled estimate of 28.5%. There was no significant time trend in case fatality rates, as assessed by logistic regression (p = 0.894). However, there was a significant negative relationship with GDP, which persisted after controlling for geographical area (the only other significant predictor in univariate analysis, likely confounded by study design). The crude OR for case fatality per $1000 increase in GDP was 0.96 (95% CI 0.92–1.00, p = 0.047), and adjusted OR = 0.86 (95% CI 0.79–0.94, p = 0.001).

Table 7. Case fatality amongst people with a diagnosis of PCP in included studies.

PCP Prophylaxis Appears Effective

Three adult studies [28], [37], [38] (total 445 patients) and four paediatric studies [21], [33], [34], [39] (total 524 patients) detailed prophylaxis use amongst patients with PCP in resource-poor settings. Table 8 details odds ratios and quality assessments for these studies. Data is not pooled due to differences in study design and populations. All studies indicated reduced odds for PCP amongst patients taking prophylaxis, although only three significantly so.

Table 8. Summary of studies detailing effects of prophylaxis on prevalence of PCP.

CD4 Count at Presentation with PCP is Generally <200×103/ml, but is Higher in LMIC than in High Income Countries

Table 9 details available data on CD4 count at presentation with PCP. Most studies report a mean or median CD4 count <200×103/ml in adults. Certain populations were documented to have higher CD4 counts: Malaysian intravenous drug users [40] (mean = 576×103/ml) and [41] (range 310-1681×103/ml), and Filipino commercial sex workers [42] (mean = 364×103/ml). However, we are unable to confirm diagnosis in these studies: two report the use of ‘sputum’, for either ‘staining’ [40] or immunofluorescence [42], but not how it was obtained. Median CD4 counts at presentation with PCP in adults in LMIC were higher than in high income countries (p = 0.028, Mann-Whitney U-test, excluding [41] as a clear outlier).

Table 9. CD4 count at presentation with Pneumocystis jirovecii pneumonia in HIV-infected people.


The most significant finding from this analysis is that the odds of a PCP diagnosis as an ADI or reason for hospitalisation in the HIV-infected population are strongly and positively correlated with per capita GDP. Although failure to make the diagnosis in resource-poor countries may contribute to this relationship (discussed below), we hypothesise that the relative scarcity of PCP in LMIC is primarily a result of poverty. In conditions of poverty, HIV-infected individuals are exposed to many infectious agents and non-infectious health challenges resulting in hospitalisation: these risks are compounded by weak health systems with poor preventative healthcare. In situations of high ‘pathogen competition’, relatively non-pathogenic organisms such as Pneumocystis will be under-represented. As countries develop economically they reduce health risks and improve the conditions underlying transmission of contagious infection: as reflected in the decreasing prevalence of tuberculosis with increasing GDP (Figure 4B). Similarly, there is a highly significant negative correlation between per capita GDP [15] and WHO age-standardized death rates (available at from meningitis or lower respiratory tract infections (Pearson correlation coefficient for meningitis -0.39, p<0.0001, n = 166 countries; Pearson correlation coefficient for respiratory infection -0.43, p<0.0001, n = 168 countries), suggesting a higher burden of severe illness with virulent pathogens in poorer countries. Pneumocystis jirovecii, ubiquitous in all settings, thus becomes a greater relative threat with economic advancement. The implicit conclusion is that HIV-infected people in wealthy countries are ‘protected’ from early presentation, since their CD4 count can drop into the range of risk for PCP before they develop illness.

As mentioned above, another explanation for the association between risk of PCP and GDP would be improving diagnostics with economic development. This may well be important, but it is worth noting that we found no relationship between diagnostic quality and GDP in these studies. Lower quality techniques did appear to reduce odds of diagnosis (although curiously the highest odds were seen when diagnostic technique was not recorded), but inclusion of diagnostic score in our model did not significantly alter the core findings. Effects of GDP were also similar when analysis was restricted to studies with high quality diagnosis, and in African clinical studies which predominantly diagnosed PCP with certainty (Table 5). Nevertheless, it remains possible that increasing awareness of PCP among clinicians results in more investigation or clinical diagnosis: this may explain some of the apparent increase over time or differences between geographical areas. However, it seems unlikely that this phenomenon would abolish the relationship we have demonstrated with GDP, since it would imply gross under-diagnosis of PCP in LMIC. In Western Europe before ART was introduced, PCP was responsible for two thirds of ADI, and 75% of HIV-infected patients developed PCP during their lifetime [43]: if PCP existed at a similar prevalence in LMIC it would certainly be identified. In contrast, high quality autopsy and clinical studies have failed to identify high rates of PCP in low income settings [2], [3], [26]. Notably, our model predicts approximately 6.5-fold increased odds of PCP diagnosis at the GDP levels of Western European countries in the early 1980s versus low-income countries (GDP<$1000 per capita), consistent with observed data.

In terms of other predictors, we observed higher odds for PCP in Central/South America compared to other areas (we are cautious about Middle Eastern results due to low numbers); higher odds in children than adults; higher odds in nationwide studies than hospitals and probably the lowest in rural areas; and an increase over time. All these factors were weaker predictors than GDP. They may reflect economic differences not reflected in national GDP figures (eg greater relative poverty in rural areas, falling ‘absolute poverty’ over time, or an effect of urbanisation). Alternatively, there may be biological explanations, most plausible for the increased risk in children. Finally, they may reflect disparities in awareness among clinicians of a condition which is difficult to diagnose.

We evaluated the challenge of clinical diagnosis by assessing features which might be suggestive of PCP in resource-poor settings. We discovered several features whose absence (cough, dyspnoea, fever) or presence (haemoptysis) pointed strongly against a diagnosis of PCP, but there was little specific to PCP itself. Although ‘typical’ chest radiographic appearances varied, the chest x-ray did provide some discriminatory power in adults. In children, all predictive features performed more poorly.

Some of this diagnostic difficulty may arise from frequent co-infection. As discussed, in situations of multiple pathogen exposure we would expect virulent organisms (highly infectious and pathogenic even in immunocompetent people) to dominate. This is borne out by high rates of M.tuberculosis and virulent bacterial infection. Correspondingly, true ‘opportunistic’ infection was not as highly represented in this data, and was often seen in high income countries, including two of three defined non-tuberculous mycobacteria (Israel [44]) and both protozoal infections (Israel [44] and Bahrain [45]). Cryptococcus was identified in 63 cases, but the majority were pulmonary infection, which does not necessarily indicate significant immunosuppression (50 cases are from a single cardio-respiratory necropsy study [46]). Although CMV was frequently detected, its significance in respiratory samples from HIV-infected patients remains uncertain [47], [48].

Virulent co-infection might be one explanation for the high case fatality from PCP in LMIC (although we do not have statistical proof for this hypothesis). Overall case fatality was 30.8% across 41 studies. This appears to be higher than in high income settings: one recent US paper reported a case fatality of 8.4% amongst 524 patients [49]. Similarly, we observed a significant negative relationship between case fatality and GDP in our data. Antimicrobial therapy against PCP seemed to be largely appropriate in these studies, but this by itself appears insufficient to prevent fatality.

This high fatality emphasises the importance of preventing PCP in these settings. Little evidence was available for ART, but more robust evidence was found for the effectiveness of prophylaxis. However, 8% of adults and 24% of children with PCP in these settings reported a history of prophylaxis receipt; use of prophylaxis also poorly predicted the absence of PCP in children (Figure 5). Practitioners should thus consider PCP even with a history of prophylaxis. Current thresholds for commencing PCP prophylaxis seem appropriate, as the majority of PCP patients in LMIC have significant cell-mediated deficiency (Table 9). However, median CD4 counts in LMIC were higher than in high income countries, and there was weak evidence that additional risk factors (intravenous drug use or commercial sex work) might predispose to PCP at higher CD4 count.

The effectiveness of ART and prophylaxis in preventing PCP potentially contradicts the negative relationship between GDP and risk of PCP, since intuitively higher income countries have greater access to these interventions. However, this is probably explained by the fact that PCP is often the presenting illness in HIV-infected individuals.


The search only included three databases, but identified a large number of studies and is therefore probably comprehensive; however, we may have missed studies with no recorded cases of PCP. Inclusion of countries further outside tropical latitude might have strengthened the relationship with GDP, although we note that PCP is three times commoner as an ADI in Western Europe than Eastern Europe [50], which is consistent with our data. Accurate population denominators were not available for the included studies, thus results indicate effects on the relative rather than absolute frequency of PCP diagnosis. However, areas differ widely in HIV prevalence, and the relative frequency of complications defines the priorities for policy-makers responsible for a defined population.


These data should alert medical practitioners and policy-makers. PCP is likely to increase in relative frequency as an HIV complication in LMIC as they develop economically, and physicians should familiarise themselves with the presentations of this infection. Prophylaxis against PCP appears effective and should be provided according to guidelines. Case fatality from PCP is high in LMIC settings, and demands improvement.

Increasing odds of PCP with rising GDP implies that wealthier economic environments allow the CD4 count to fall further than in LMIC before HIV infection presents. This emphasises the importance in all settings of identifying HIV infection early. Those presenting with PCP usually have a CD4 count <200×103/ml, and low nadir CD4 predicts poorer long term prognosis from HIV infection [51].

In summary, we have demonstrated a link between GDP and diagnosis of PCP, and thereby provide an explanation for the previously-described epidemiology in tropical and low income settings.

Supporting Information

References S1. Additional references reviewed and analysed but not specifically cited in the manuscript.


Author Contributions

Conceived and designed the experiments: DML MXR CDJ RFM. Performed the experiments: DML RFM. Analyzed the data: DML FG. Wrote the paper: DML MXR CDJ RFM.


  1. 1. Huang L, Cattamanchi A, Davis JL, Boon SD, Kovacs J, et al. (2011) HIV-Associated Pneumocystis Pneumonia. Proc Am Thorac Soc 8(3): 294–300.
  2. 2. Lucas SB, Peacock CS, Hounnou A, Brattegaard K, Koffi K, et al. (1996) Disease in children infected wth HIV in Abidjan, Cote d’Ivoire. BMJ 312(7027): 335–8.
  3. 3. Abouya YL, Beaumel A, Lucas S, Dago-Akribi A, Coulibaly G, et al. (1992) Pneumocystis carinii pneumonia. An uncommon cause of death in African patients with acquired immunodeficiency syndrome. Am Rev Respir Dis 145(3): 617–20.
  4. 4. Smulian AG, Sullivan DW, Linke MJ, Halsey NA, Quinn TC, et al. (1993) Geographic variation in the humoral response to Pneumocystis carinii. J Infect Dis 167(5): 1243–7.
  5. 5. Daly KR, Fichtenbaum CJ, Tanaka R, Linke MJ, O’Bert R, et al. (2002) Serologic responses to epitopes of the major surface glycoprotein of Pneumocystis jiroveci differ in human immunodeficiency virus-infected and uninfected persons. J Infect Dis 186(5): 644–51.
  6. 6. Respaldiza N, Medrano FJ, Medrano AC, Varela JM, de la Horra C, et al. (2004) High seroprevalence of Pneumocystis infection in Spanish children. Clin Microbiol Infect 10(11): 1029–31.
  7. 7. Nkinin SW, Daly KR, Walzer PD, Ndzi ES, Asonganyi T, et al. (2009) Evidence for high prevalence of Pneumocystis jirovecii exposure among Cameroonians. Acta Trop 112(2): 219–24.
  8. 8. Miller RF, Lindley AR, Malin AS, Ambrose HE, Wakefield AE (2005) Isolates of Pneumocystis jirovecii from Harare show high genotypic similarity to isolates from London at the superoxide dismutase locus. Trans R Soc Trop Med Hyg 99(3): 202–6.
  9. 9. Hu DJ, Fleming PL, Castro KG, Jones JL, Bush TJ, et al. (1995) How important is race/ethnicity as an indicator of risk for specific AIDS-defining conditions? J Acquir Immune Defic Syndr Hum Retrovirol 10(3): 374–80.
  10. 10. Parkhurst JO (2010) Understanding the correlations between wealth, poverty and human immunodeficiency virus infection in African countries. Bull World Health Organ 88(7): 519–26.
  11. 11. Janssens J-P, Rieder HL (2008) An ecological analysis of incidence of tuberculosis and per capita gross domestic product. Eur Respir J 32(5): 1415–6.
  12. 12. Yazdanpanah Y, Chêne G, Losina E, Goldie SJ, Merchadou LD, et al. (2001) Incidence of primary opportunistic infections in two human immunodeficiency virus-infected French clinical cohorts. Int J Epidemiol 30(4): 864–71.
  13. 13. Marras TK, Daley CL (2002) Epidemiology of human pulmonary infection with nontuberculous mycobacteria. Clin Chest Med 23(3): 553–67.
  14. 14. Thomas CF Jr, Limper AH (2004) Pneumocystis pneumonia. N Engl J Med 350(24): 2487–98.
  15. 15. World Bank Indicators. Available: Accessed 3 October 2012.
  16. 16. Prisma Statement. Available: Accessed: 3 October 2012.
  17. 17. Machiels G, Urban MI (1992) Pneumocystis carinii as a cause of pneumonia in HIV-infected patients in Lusaka, Zambia. Trans R Soc Trop Med Hyg 86(4): 399–400.
  18. 18. Siika AM, Chakaya JM, Revathi G, Mohamed SS, Bhatt KM (2006) Bronchoscopic study on aetiology of chronic cough in HIV-infected adults with negative sputum smears for Mycobacterium tuberculosis at Kenyatta National Hospital, Nairobi. East Afr Med J 83(6): 295–305.
  19. 19. Rupali P, Abraham OC, Zachariah A, Subramanian S, Mathai D (2003) Aetiology of prolonged fever in antiretroviral-naïve human immunodeficiency virus-infected adults. Natl Med J India 16(4): 193–9.
  20. 20. Senya C, Mehta A, Harwell JI, Pugatch D, Flanigan T, et al. (2003) Spectrum of opportunistic infections in hospitalized HIV-infected patients in Phnom Penh, Cambodia. Int J STD AIDS 14(6): 411–6.
  21. 21. Madhi SA, Cutland C, Ismail K, O’Reilly C, Mancha A, et al. (2002) Ineffectiveness of trimethoprim-sulfamethoxazole prophylaxis and the importance of bacterial and viral coinfections in African children with Pneumocystis carinii pneumonia. Clin Infect Dis 35(9): 1120–6.
  22. 22. Lian YLA, Heng BSL, Nissapatorn V, Lee C (2007) AIDS-defining illnesses: a comparison between before and after commencement of highly active antiretroviral therapy (HAART). Curr HIV Res 5(5): 484–9.
  23. 23. Kouakoussui A, Fassinou P, Anaky MF, Elenga N, Laguide R, et al. (2004) Respiratory manifestations in HIV-infected children pre- and post-HAART in Abidjan, the Ivory Coast. Paediatr Resp Rev 5(4): 311–5.
  24. 24. Saraceni V, King BS, Cavalcante SC, Golub JE, Lauria LM, et al. (2008) Tuberculosis as primary cause of death among AIDS cases in Rio de Janeiro, Brazil. Int J Tuberc Lung Dis 12(7): 769–72.
  25. 25. Chariyalertsak S, Sirisanthana T, Saengwonloey O, Nelson KE (2001) Clinical presentation and risk behaviors of patients with acquired immunodeficiency syndrome in Thailand, 1994–1998: regional variation and temporal trends. Clin Infect Dis 32(6): 955–62.
  26. 26. Batungwanayo J, Taelman H, Lucas S, Bogaerts J, Alard D, et al. (1994) Pulmonary disease associated with the human immunodeficiency virus in Kigali, Rwanda. A fiberoptic bronchoscopic study of 111 cases of undetermined etiology. Am J Respir Crit Care Med 149(6): 1591–6.
  27. 27. Hargreaves NJ, Kadzakumanja O, Phiri S, Lee CH, Tang X, et al. (2001) Pneumocystis carinii pneumonia in patients being registered for smear-negative pulmonary tuberculosis in Malawi. Trans R Soc Trop Med Hyg 95(4): 402–8.
  28. 28. Le Minor O, Germani Y, Chartier L, Lan NH, Lan NT, et al. (2008) Predictors of pneumocystosis or tuberculosis in HIV-infected Asian patients with AFB smear-negative sputum pneumonia. J Acquir Immune Defic Syndr 48(5): 620–7.
  29. 29. Malin AS, Gwanzura LKZ, Klein S, Robertson VJ, Musvaire P, et al. (1995) Pneumocystis carinii pneumonia in Zimbabwe. Lancet 346(8985): 1258–61.
  30. 30. Mootsikapun P, Chetchotisakd P, Intarapoka B (1996) Pulmonary infections in HIV infected patients. J Med Assoc Thai 79(8): 477–85.
  31. 31. Tansuphasawadikul S, Pitisuttithum P, Knauer AD, Supanaranond W, Kaewkungwal J, et al. (2005) Clinical features, etiology and short term outcomes of interstitial pneumonitis in HIV/AIDS patients. Southeast Asian J Trop Med Public Health 36(6): 1469–78.
  32. 32. Chakaya JM, Bii C, Ng’ang’a L, Amukoye E, Ouko T, et al. (2003) Pneumocystis carinii pneumonia in HIV/AIDS patients at an urban district hospital in Kenya. East Afr Med J 80(1): 30–5.
  33. 33. Chokephaibulkit K, Wanachiwanawin D, Chearskul S, Wanprapa N, Unganont K, et al. (1999) Pneumocystis carinii severe pneumonia among human immunodeficiency virus-infected children in Thailand: the effect of a primary prophylaxis strategy. Pediatr Infect Dis J 18(2): 147–52.
  34. 34. Ruffini DD, Madhi SA (2002) The high burden of Pneumocystis carinii pneumonia in African HIV-1-infected children hospitalized for severe pneumonia. AIDS 16(1): 105–12.
  35. 35. Zar HJ, Dechaboon A, Hanslo D, Apolles P, Magnus KG, et al. (2000) Pneumocystis carinii pneumonia in South African children infected with human immunodeficiency virus. Pediatr Infect Dis J 19(7): 603–7.
  36. 36. Graham SM, Mtitimila EI, Kamanga HS, Walsh AL, Hart CA, et al. (2000) Clinical presentation and outcome of Pneumocystis carinii pneumonia in Malawian children. Lancet 355(9201): 369–73.
  37. 37. Anekthananon T, Ratanasuwan W, Techasathit W, Rongrungruang Y, Suwanagool S (2004) HIV infection/acquired immunodeficiency syndrome at Siriraj Hospital, 2002: time for secondary prevention. J Med Assoc Thai 87(2): 173–9.
  38. 38. Knauer A, Das AK, Tansuphasawadikul S, Supanaranond W, Pitisuttithum P, et al.. (2005) Clinical features, aetiology and short-term outcome of interstitial pneumonitis in HIV/AIDS patients at Bamrasnaradura hospital, Nonthaburi, Thailand. Wien Klin Wochenschr (Suppl 4): 49–55.
  39. 39. Zar HJ, Hanslo D, Tannenbaum E, Klein M, Argent A, et al. (2001) Aetiology and outcome of pneumonia in human immunodeficiency virus-infected children hospitalized in South Africa. Acta Paediatr 90(2): 119–25.
  40. 40. Wood E, Cheong I, Lee C (1998) A study of chest infections in HIV seropositive patients in Kuala Lumpur. Int J Clin Pract 52(1): 23–6.
  41. 41. Yoong KY, Cheong I (1997) A study of Malaysian drug addicts with human immunodeficiency virus infection. Int J STD AIDS 8(2): 118–23.
  42. 42. Manaloto CR, Perrault JG, Caringal LT, Santiago EG, Wignall FS, et al. (1994) Natural history of HIV infection in Filipino female commercial sex workers. J Acquir Immune Defic Syndr 7(11): 1157–68.
  43. 43. Morris A, Lundgren JD, Masur H, Walzer PD, Hanson DL, et al. (2004) Current epidemiology of Pneumocystis pneumonia. Emerg Infect Dis 10(10): 1713–20.
  44. 44. Siegman-Igra Y, Maayan S, Pitlik SD, Costin C, Swartz T, et al. (1988) AIDS in Israel, 1987. Isr J Med Sci 24(3): 131–6.
  45. 45. al-Haddad MK, Baig BZ, Ebrahim RA (1997) Epidemiology of HIV and AIDS in Bahrain. J Commun Dis 29(4): 321–8.
  46. 46. Wong ML, Back P, Candy G, Nelson G, Murray J (2006) Pneumocystis jiroveci pneumonia in African miners at autopsy. Int J Tuberc Lung Dis 10(7): 756–60.
  47. 47. de la Hoz RE, Stephens G, Sherlock C (2002) Diagnosis and treatment approaches of CMV infections in adult patients. J Clin Virol (Suppl 2): S1–12.
  48. 48. Huang L, Crothers K (2009) HIV-associated opportunistic pneumonias. Respirology 14(4): 474–85.
  49. 49. Fei MW, Kim EJ, Sant CA, Jarlsberg LG, Davis JL, et al. (2009) Predicting mortality from HIV-associated Pneumocystis pneumonia at illness presentation: an observational cohort study. Thorax 64(12): 1070–6.
  50. 50. Serraino D, Puro V, Boumis E, Angeletti C, Girardi E, et al. (2003) Epidemiological aspects of major opportunistic infections of the respiratory tract in persons with AIDS: Europe, 1993–2000. AIDS 17(14): 2109–16.
  51. 51. When To Start Consortium, Sterne JA, May M, Costagliola D, de Wolf F, et al. (2009) Timing of initiation of antiretroviral therapy in AIDS-free HIV-1-infected patients: a collaborative analysis of 18 HIV cohort studies. Lancet 373(9672): 1352–63.
  52. 52. Moreira Júnior ED, Silva N, Brites C, Carvalho EM, Bina JC, et al. (1993) Characteristics of the acquired immunodeficiency syndrome in Brazil. Ann J Trop Med Hyg 48(5): 687–92.
  53. 53. Udwadia ZF, Doshi AV, Bhaduri AS (2005) Pneumocystis carinii pneumonia in HIV infected patients from Mumbai. J Assoc Physicians India 53: 437–40.
  54. 54. Sharma SK, Kadhiravan T, Banga A, Goyal T, Bhatia I, et al. (2004) Spectrum of clinical disease in a series of 135 hospitalised HIV-infected patients from north India. BMC Infect Dis 4: 52.
  55. 55. Bellamy R, Sangeetha S, Paton NI (2004) AIDS-defining illnesses among patients with HIV in Singapore, 1985 to 2001: results from the Singapore HIV Observational Cohort Study (SHOCS). BMC Infect Dis 4: 47.
  56. 56. Nissapatorn V, Lee C, Fatt QK, Abdullah KA (2003) AIDS-related opportunistic infections in Hospital Kuala Lumpur. Jpn J Infect Dis 56(5–6): 187–92.
  57. 57. Singh A, Bairy I, Shivananda PG (2003) Spectrum of opportunistic infections in AIDS cases. Indian J Med Sci 57(1): 16–21.
  58. 58. Kim JM, Cho GJ, Hong SK, Chang KH, Chung JS, et al. (2003) Epidemiology and clinical features of HIV infection/AIDS in Korea. Yonsei Med J 44(3): 363–70.
  59. 59. Inverarity D, Bradshaw Q, Wright P, Grant A (2002) The spectrum of HIV-related disease in rural Central Thailand. Southeast Asian J Trop Med Public Health 33(4): 822–31.
  60. 60. Kumarasamy N, Solomon S, Flanigan TP, Hemalatha R, Thyagarajan SP, et al. (2003) Natural history of human immunodeficiency virus disease in Southern India. Clin Infect Dis 36(1): 79–85.
  61. 61. Louie JK, Chi NH, Thao LTT, Quang VM, Campbell J, et al. (2004) Opportunistic infections in hospitalized HIV-infected adults in Ho Chi Minh City, Vietnam: a cross-sectional study. Int J STD AIDS 15(11): 758–61.
  62. 62. Fang CT, Hung CC, Chang SC, Hsueh PR, Chang YL, et al. (2000) Pulmonary infection in human immunodeficiency virus-infected patients in Taiwan. J Formos Med Assoc 99(2): 123–7.
  63. 63. Oh MD, Park SW, Kim HB, Kim NJ, Choi HJ, et al. (1999) Spectrum of opportunistic infections and malignancies in patients with human immunodeficiency virus infection in South Korea. Clin Infect Dis 29(6): 1524–8.
  64. 64. Tansuphasawadikul S, Amornkul PN, Tanchanpong C, Limpakarnjanarat K, Kaewkungwal J, et al. (1999) Clinical presentation of hospitalized adult patients with HIV infection and AIDS in Bangkok, Thailand. J Acquir Immune Defic Syndr 21(4): 326–32.
  65. 65. George J, Hamide A, Das AK, Amarnath SK, Rao RS (1996) Clinical and laboratory profile of sixty patients with AIDS: a South Indian study. Southeast Asian J Trop Med Public Health 27(4): 686–91.
  66. 66. Nissapatorn V, Lee CKC, Rohela M, Anuar AK (2004) Spectrum of opportunistic infections among HIV-infected patients in Malaysia. Southeast Asian J Trop Med Public Health (Suppl 2): 26–32.
  67. 67. Santos B, Beck EJ, Peixoto MF (1994) Survival and medical intervention in southern Brazilian AIDS patients. Int J STD AIDS 5(4): 279–83.
  68. 68. Tan BH, Wong GC, Lam MS, Ang BS, Wong SY (1997) Clinical characteristics and natural history of human immunodeficiency virus infected patients seen at a general hospital in Singapore. Ann Acad Med Singapore 26(5): 566–74.
  69. 69. Swasdisevi A (1994) Clinical study of HIV disease in the lower area of northern Thailand in 1994. J Med Assoc Thai 77(8): 440–3.
  70. 70. Vithayasai P, Vithayasai V (1993) Clinical manifestations of 174 AIDS cases in Maharaj Nakorn Chiang Mai Hospital. J Dermatol 20(7): 389–93.
  71. 71. Kong BN, Harmwell JI, Suos P, Lynen L, Mohiuddin S, et al. (2007) Opportunistic infections and HIV clinical disease stage among patients presenting for care in Phnom Penh, Cambodia. Southeast Asian J Trop Med Public Health 38(1): 62–8.
  72. 72. Wong KH, Lee SS (1998) Comparing the first and second hundred AIDS cases in Hong Kong. Singapore Med J 39(6): 236–40.
  73. 73. Shah I (2005) Age related clinical manifestations of HIV infection in Indian children. J Trop Pediatr 51(5): 300–3.
  74. 74. Rajasekaran S, Jeyaseelan L, Raja K, Ravichandran N (2009) Demographic & clinical profile of HIV infected children accessing care at Tambaram, Chennai, India. Indian J Med Res 129(1): 42–9.
  75. 75. Solomon SS, Kumarasamy N, Celentano DD, Yepthomi TH, Arvind VP, et al. (2006) Trends in HIV-related morbidity among patients admitted to a South Indian tertiary hospital between 1997 and 2003. AIDS Care 18(4): 366–70.
  76. 76. Soares VY, Lúcio Filho CE, Carvalho LI, Silva AM, Eulálio KD (2008) Clinical and epidemiological analysis of patients with HIV/AIDS admitted to a reference hospital in the northeast region of Brazil. Rev Inst Med Trop Sao Paulo 50(6): 327–32.
  77. 77. Bedri A, Lulseged S (2001) Clinical description of children with HIV/AIDS admitted at a referral hospital in Addis Ababa. Ethiop Med J 39(3): 203–11.
  78. 78. Zar HJ, Apolles P, Argent A, Klein M, Burgess J, et al. (2001) The etiology and outcome of pneumonia in human immunodeficiency virus-infected children admitted to intensive care in a developing country. Pediatr Crit Care Med 2(2): 108–112.
  79. 79. Kay-Thwe-Han Rai-Mra, Htin-Aung-Saw Ye-Htut (2003) Wynn-Naing (2003) Pneumocystis carinii infection among human immunodeficiency virus (HIV) infected Myanmar patients. Southeast Asian J Trop Med Public Health 34(3): 577–9.
  80. 80. van Oosterhout JJ, Laufer MK, Graham SM, Thumba F, Perez MA, et al. (2005) A community-based study of the incidence of trimethoprim-sulfamethoxazole-preventable infections in Malawian adults living with HIV. J Acquir Immune Defic Syndr 39(5): 626–31.
  81. 81. van Oosterhout JJG, Laufer MK, Arantza Perez M, Graham SM, Chimbiya N, et al. (2007) Pneumocystis pneumonia in HIV-positive adults, Malawi. Emerg Inf Dis 13(2): 325–8.
  82. 82. Aderaye G, Bruchfeld J, Aseffa G, Nigussie Y, Melaku K, et al. (2007) Pneumocystis jiroveci pneumonia and other pulmonary infections in TB smear-negative HIV-positive patients with atypical chest X-ray in Ethiopia. Scand J Infect Dis 39(11–12): 1045–53.
  83. 83. Kibiki GS, Beckers P, Mulder B, Arens T, Mueller A, et al. (2007) Aetiology and presentation of HIV/AIDS-associated pulmonary infections in patients presenting for bronchoscopy at a referral hospital in northern Tanzania. East Afr Med J 84(9): 420–8.
  84. 84. Giri TK, Pande I, Mishra NM, Kailash S, Uppal SS, et al. (1995) Spectrum of clinical and laboratory characteristics of HIV infection in northern India. J Commun Dis 27(3): 131–41.
  85. 85. Morrow BM, Hsaio NY, Zampoli M, Whitelaw A, Zar HJ (2010) Pneumocystis pneumonia in South African children with and without human immunodeficiency virus infection in the era of highly active antiretroviral therapy. Pediatr Infect Dis J 29(6): 535–9.
  86. 86. Hung CC, Chen MY, Hsieh SM, Sheng WH, Chang SC (2000) Clinical spectrum, morbidity, and mortality of acquired immunodeficiency syndrome in Taiwan: a 5-year prospective study. J Acquir Immune Defic Syndr 24(4): 378–85.
  87. 87. Moses AE, Maayan S, Rahav G, Weinberger M, Engelhard D, et al. (1996) HIV infection and AIDS in Jerusalem: a microcosm of illness in Israel. Isr J Med Sci 32(9): 716–21.
  88. 88. Atzori C, Bruno A, Chichino G, Gatti S, Scaglia M (1993) Pneumocystis carinii pneumonia and tuberculosis in Tanzanian patients infected with HIV. Trans R Soc Trop Med Hyg 87(1): 55–6.
  89. 89. Corbett EL, Churchyard GJ, Charalambos S, Samb B, Moloi V, et al. (2002) Morbidity and mortality in South African gold miners: impact of untreated disease due to human immunodeficiency virus. Clin Infect Dis 34(9): 1251–8.
  90. 90. Carme B, Mboussa J, Andzin M, Mbouni E, Mpele P, et al. (1991) Pneumocystis carinii is rare in AIDS in Central Africa. Trans R Soc Trop Med Hyg 85(1): 80.
  91. 91. Delport SD, Brisley T (2002) Aetiology and outcome of severe community-acquired pneumonia in children admitted to a paediatric intensive care unit. S Afr Med J 92(11): 907–11.
  92. 92. Bakeera-Kitaka S, Musoke P, Downing R, Tumwine JK (2004) Pneumocystis carinii in children with severe pneumonia at Mulago Hospital, Uganda. Ann Trop Paediatr 24(3): 227–35.
  93. 93. Kamiya Y, Mtitimila E, Graham SM, Broadhead RL, Brabin B, et al. (1997) Pneumocystis carinii pneumonia in Malawian children. Ann Trop Paediatr 17(2): 121–6.
  94. 94. Uriyo J, Gosling RD, Maddox V, Sam NE, Schimana W, et al. (2006) Prevalences of Pneumocystis jiroveci, Mycobacterium tuberculosis and Streptococcus pnrumoniae infection in children with severe pneumonia, in a tertiary referral hospital in northern Tanzania. Ann Trop Med Parasitol 100(3): 245–9.
  95. 95. Rabie H, de Boer A, van den Bos S, Cotton MF, Kling S, et al. (2007) Children with human immunodeficiency virus infection admitted to a paediatric intensive care unit in South Africa. J Trop Pediatr 53(4): 270–3.
  96. 96. Narata R, Wangkaew S, Kasitanon N, Louthrenoo W (2007) Community-acquired pneumonia in Thai patients with systemic lupus erythematosus. Southeast Asian J Trop Med Public Health 38(3): 528–36.
  97. 97. Ismail R, Doi S, Naganathna N (1995) HIV infection in Malaysia: a report of cases seen at the University Hospital, Kuala Lumpur. Med J Malaysia 50(4): 298–301.
  98. 98. Lee CC, Leo YS, Snodgrass I, Wong SY (1997) The demography, clinical manfestations and natural history of human immunodeficiency virus (HIV) infection in an older population in Singapore. Ann Acad Med Singapore 26(6): 731–5.
  99. 99. Weinberg A, Duarte MIS (1993) Respiratory complications in Brazilian patients infected with human immunodeficiency virus. Rev Inst Med Trop Sao Paulo 34(6): 129–39.
  100. 100. Lambertucci JR, Rayes AAM, Nunes F, Landazuri-Palacios JE, Nobre V (1999) Fever of undetermined origin in patients with the acquired immunodeficiency syndrome in Brazil: report on 55 cases. Rev Inst Med Trop S Paulo 41(1): 27–32.
  101. 101. Fallo AA, DobrzanskiNisiewicz W, Sordelli N, Cattaneo MA, Scott G, et al. (2002) Clinical and epidemiologic aspects of human immunodeficiency virus-1-infected children in Buenos Aires, Argentina. Int J Infect Dis 6(1): 9–16.
  102. 102. Arora VK, Kumar SV (1999) Pattern of opportunistic pulmonary infections in HIV sero-positive subjects: observations from Pondicherry, India. Indian J Chest Dis Allied Sci 41(3): 135–44.
  103. 103. Usha MM, Rajendran P, Thyagarajan SP, Solomon S, Kumarasamy N, et al. (2000) Identification of Pneumocystis carinii in induced sputum of AIDS patients in Chennai (Madras). Indian J Pathol Microbiol 43(3): 291–6.
  104. 104. Merchant RH, Oswal JS, Bhagwat RV, Karkare J (2001) Clinical profile of HIV infection. Indian Pediatr 38(3): 239–46.
  105. 105. Madhivanan P, Mothi SN, Kumarasamy N, Yepthomi T, Venkatesan C, et al. (2003) Clinical manifestations of HIV infected children. Indian J Pediatr 70(8): 615–20.