Abstract
The microbiologic etiology of severe pneumonia in hospitalized patients is rarely known in sub-Saharan Africa. Through a comprehensive diagnostic work-up, we aimed to identify the causative agent in severely ill patients with a clinical picture of pneumonia admitted to a high-dependency unit. A final diagnosis was made and categorized as confirmed or probable by using predefined criteria. Fifty-one patients were recruited (45% females), with a mean age of 35 years (range = 17–88 years), of whom 11(22%) died. Forty-eight (94%) of the patients were seropositive for human immunodeficiency virus; 14 (29%) of these patients were receiving antiretroviral treatment. Final diagnoses were bacterial pneumonia (29%), Pneumocystis jirovecii pneumonia (27%), pulmonary tuberculosis (22%), and pulmonary Kaposi's sarcoma (16%); 39 (77%) of these cases were confirmed cases. Fifteen (29%) patients had multiple isolates. At least 3 of 11 viral-positive polymerase chain reaction (PCR) results of bronchoalveolar lavage fluid were attributed clinical relevance. No atypical bacterial organisms were found.
Introduction
Severe pulmonary infections in sub-Saharan African adults are common and frequently lead to hospitalization.1 Because of the unavailability of expensive and technically demanding diagnostic facilities, the etiology often remains unclear, and antimicrobial treatment is therefore empirical rather than being directed against an identified pathogen. Such a clinically based diagnosis lacks accuracy and carries the risk of inadequate antibiotic treatment, in particular because physical signs and radiographic manifestations in advanced immunodeficiency and the immune reconstitution inflammatory syndrome (IRIS) are commonly insensitive and non-specific.2–4 In the few reported studies on the clinical and microbiologic spectrum of lower respiratory infections in adults in sub-Saharan Africa, bacterial pneumonia, pulmonary tuberculosis (PTB), and Pneumocystis jirovecii pneumonia (PcP) were among the most common causes but studies differed in methods used and the diagnostic work-up was frequently non-exhaustive.1,5–8 Even fewer data are available, especially from patients with the most severe pneumonia, on the causative roles of viral, fungal, and atypical bacterial pathogens, and on the contribution of Kaposi's sarcoma (KS) to apparently pneumonic syndromes.1,8
Pulmonary tuberculosis and pneumococcal pneumonia are frequent diagnoses in patients admitted to the Queen Elizabeth Central Hospital (QECH) in Blantyre, Malawi.9–11 The importance of Pneumocystis jirovecii pneumonia has been described for hospitalized children and in adult community and outpatient-based settings in Malawi but is less clear in hospitalized adults.12–14 In 2004, we set up a medical high dependency unit (MHDU) at QECH to improve care for severely ill adult medical patients, in particular those with respiratory distress and hypoxia. An audit of patients admitted to this MHDU showed that under routine circumstances with limited diagnostic capacity, microbiologic confirmation could not be achieved in any of the suspected PcP cases and in only in 6 of 40 bacterial pneumonia cases, thus frequently resulting in diagnostic and therapeutic uncertainty (Ajdukiewicz KM, Zijlstra EE, unpublished data).
We therefore conducted a prospective study carried out at a time of ongoing antiretroviral therapy (ART) scale-up in Malawi by using a comprehensive diagnostic work-up to determine final diagnoses in severely ill patients who were admitted to a MHDU with a clinical picture of pneumonia.
Materials and Methods
Setting, patients, and case definition.
The study was carried out at QECH in Blantyre, the largest health care facility in the country, which serves approximately one million persons. Approximately 70% of adult medical in-patients are infected with human immunodeficiency virus (HIV) and 40% have acquired immunodeficiency syndrome (AIDS).15 The hospital has a six-bed MHDU that is equipped with oxygen concentrators providing a maximum flow of 5 liters/minute.
All patients admitted to the MHDU during February–September 2006 were screened for enrollment. Inclusion criteria were an age ≥ 18 years, at least one sample of expectorated sputum examined that was negative for acid-fast bacilli (AFB), and a clinical diagnosis of pneumonia severe enough to warrant admission to MHDU for supplemental oxygen. The case definition of pneumonia entailed at least one symptom of cough, sputum production, breathlessness, chest pain, and hemoptysis, and an abnormality on a chest radiograph consistent with infection and prescription of empirical antimicrobial therapy for suspected pneumonia by the admitting clinician.
A person was excluded from the study if the diagnosis on admission was not pneumonia, if written informed consent could not be obtained, or if the patient was unfit for bronchoscopy. Patients were followed-up until they were discharged from the hospital.
Investigations and laboratory methods.
All samples were processed and analyzed at the Malawi-Liverpool Wellcome Trust research laboratories in Blantyre unless stated otherwise. An expectorated sputum sample was examined with modified Ziehl-Neelsen stain for AFB. Venous blood samples were obtained at admission for aerobic culture (BacTec, Becton-Dickinson, Sparks, MD), complete blood count (automated HmX; Beckman Coulter, Brea, CA), HIV test (2 rapid tests: Uni-Gold; Trinity Biotech, Wicklow, Ireland and Determine; Abbott Laboratories, Abbott Park, IL, conducted at the QECH laboratory), and CD4 count (fluorescein-activated cell sorter count; Becton Dickinson, conducted at the QECH laboratory). Flexible bronchoscopy (Olympus, Center Valley, PA) and bronchoalveolar lavage (BAL) were carried out as soon as feasible after admission and if peripheral oxygen saturation measured by pulse-oxymetry was > 90% with supplemental oxygen.16 The endoscope was unprotected for bacterial contamination when passing through the nasopharynx. A maximum of 200 mL of sterile water in aliquots of 50 mL was instilled into the affected lobar bronchus or, in diffuse disease, into the right middle lobe bronchus, and re-aspirated to the extend possible, to provide a BAL sample.
The BAL fluid was used for Pneumocystis indirect immuno-fluorescence testing (Detect IF; Axis-Shield Diagnostics Ltd., Dundee, Scotland). Bacterial, fungal, and mycobacterial cultures were set up on blood and chocolate agar plates (incubated at 37°C in a CO2 incubator for 48 hours), cystine-lactose electrolyte deficient agar plates (incubated aerobically for 24 hours), Sabouraud's agar plates (incubated aerobically for up to 2 weeks) and Lowenstein-Jensen slopes (incubated aerobically at 37°C for 10 weeks).
A single slide was prepared for modified Ziehl-Neelsen staining, and 2 mL of BAL fluid was stored at –20°C. This specimen was used for a polymerase chain reaction (PCR) of 15 respiratory viral pathogens including influenza A and B virus; parainfluenza virus type 1–4; adenovirus; rhinovirus; respiratory syncytial virus type A and B; human metapneumovirus; coronavirus 229E and OC43; human coronavirus NL; bocavirus (Department of Virology, Erasmus Medical Center, Rotterdam, Netherlands); and Pneumocystis jirovecii, Mycoplasma pneumoniae, Chlamydophila pneumoniae, Legionella pneumophila, and Chlamydophila psittaci (Department of Medical Microbiology, Leiden University Medical Center, Leiden, The Netherlands). A detailed description of the PCR methods used has been reported.17,18 Real-time PCR analysis enabled quantification of Pneumocystis jirovecii. A threshold cycle (Ct) value ≤ 35 has been associated with infection and a Ct value > 35 indicates colonization.18,19 The PCR and mycobacterial culture results were received after the study was completed and therefore did not influence patient management.
Outcome.
A final diagnosis was established retrospectively for each patient after reviewing all available clinical and microbiologic data. The final confirmed and probable diagnoses definitions are shown in Table 1. Clinical features dictated which final diagnosis was considered primary, if multiple diagnoses were made in a patient. When detected, non-pathogenic bacteria such as alpha-hemolytic streptococci and Staphylococcus epidermidis were reported as contaminants.
Table 1.
Disease | Confirmed | Probable |
---|---|---|
Bacterial pneumonia | - Positive culture of BAL and/or blood | - Suggestive clinical picture and improvement after receiving antibiotics |
- Positive BAL PCR result for atypical bacteria | ||
Pneumocystis pneumonia | - Positive BAL immuno-fluorescence | - Positive BAL PCR result (cycle threshold value < 35)18,19 and suggestive clinical picture and improvement after treatment for PcP |
(> 5 oocysts per high-power field) | ||
Pulmonary tuberculosis | - Positive BAL mycobacterial culture and/or AFB ZN smear | - Suggestive clinical picture and improvement after treatment for tuberculosis |
Pulmonary Kaposi's sarcoma | - Characteristic tracheo-bronchial lesions on bronchoscopy and suggestive chest radiograph | |
Viral pneumonitis | - Positive BAL viral PCR result and suggestive clinical picture |
BAL = bronchoalveolar lavage; PCR = polymerase chain reaction; PCP = Pneumocystis jirovecii pneumonia; AFB = acid-fast bacilli; ZN = Ziehl-Neelsen.
Case management.
Patients received empirical antibiotic treatment for pneumonia and sepsis, which depending on availability, included benzylpenicillin, ampicillin, amoxicillin, third-generation cephalosporins, chloramphenicol, macrolides, or ciprofloxacin. Patients with PcP were treated with high dose co-trimoxazole (trimethoprim/sulfamethoxazole) together with oral corticosteroids, and those considered to have PTB were started on standard four-drug anti-tuberculous treatment. Patients with pulmonary Kaposi's sarcoma (PKS) were assessed by the hospital palliative care team and received vincristine mono-chemotherapy and high-dose corticosteroids if deemed appropriate. Administration of oxygen by nasal prongs aimed to maintain peripheral oxygen saturation above 90%. In agreement with hospital policy, commencement of ART was deferred until after stabilization and discharge from hospital. Patients who were already receiving ART continued this therapy during hospitalization.
Ethical approval.
Ethical approval for the study was received from the College of Medicine Research and Ethics Committee, Blantyre, Malawi. This committee is formally mandated by the National Research Council of Malawi to review proposals emanating from or linked to the College of Medicine. Written informed consent was obtained from all patients.
Statistical analysis.
Statistical analysis was carried out by using SPSS 11.0 (SPSS, Chicago, IL) for windows. Categorical data were compared by using the chi-square test. Continuous data were compared by using t tests for measures that were normally distributed and the Mann-Whitney test for measures that were skewed in distribution (as assessed by Kolmogorov-Smirnov statistics). Levene's test for equality of variances was performed to ensure homogeneity of variances. A P value < 0.05 indicated statistical significance. Adjustment of blood hemoglobin concentration for sex was performed by binary logistic regression analysis.
Results
Patient characteristics, clinical features, and mortality.
During the 28-week study period, 159 patients were admitted to the MHDU, of whom 51 were recruited as shown in Table 2. A total of 45% of the study participants were women (mean age = 35 years, range = 17–88 years).
Table 2.
No. patients | Reasons for exclusion |
---|---|
48 | Non-respiratory diagnoses* |
19 | Non-pneumonic respiratory diagnoses† |
8 | Sputum smear-positive pulmonary tuberculosis |
33 | Pneumonic diagnoses but unsuitable for bronchoscopy‡ |
51 | Recruited |
Non-respiratory diagnoses were cardiac disease (18), meningitis (6), tuberculous pericarditis (5), severe anemia (2), pyelonephritis (2), and other (15).
Non-pneumonic respiratory diagnoses were extra pulmonary tuberculosis (10), chronic airways disease (3), pleural sepsis (3), bronchiectasis (1), and malignancy (2).
Unsuitable for bronchoscopy due to early death < 24 hours (15), hypoxia < 90% (6), confusion (2), refusal (2), cough (1), unavailability of staff (6), and under age (1).
At admission, 43 (84%) patients were presumed to have a bacterial pneumonia and 7 (14%) had Pneumocystis jirovecii pneumonia on the basis of clinical impression alone. Most (66%) patients had symptoms of at least three weeks duration (Table 3). A total of 94% of the participants were infected with HIV, of whom 29% were receiving ART that had been started less than 2 months before admission in 71% of the patients. The mean CD4 count at admission was not significantly different between HIV-infected persons who were receiving ART and those who were not receiving ART (135 versus 116 cells/μL; P = 0.77). The mean duration of hospitalization was shorter in those receiving ART than in those not receiving ART (11 versus 16 days; P = 0.05) but survival rates were similar. Nine patients died during the MHDU stay after a mean duration of stay of 8 days (range = 4–16 days). An additional two 2 inpatient deaths occurred after discharge from the MHDU, one caused by cryptococcal meningitis and one caused by suspected wasting syndrome and severe anemia. Inpatient mortality was not associated with any of the baseline characteristics, clinical features, or blood results, except hemoglobin concentration, which was higher in survivors.
Table 3.
Characteristic | All (n = 51) | Alive (n = 40) | Dead (n = 11) | OR (95% CI)† | |
---|---|---|---|---|---|
Female | 23 (39%) | 19 (48%) | 4 (36%) | 1.58 (0.40–6.27) | |
HIV status known at study entry | 31 (61%) | 23 (58%) | 8 (73%) | 0.50 (0.11–2.20) | |
HIV positive | 48 (94%) | 37 (93%) | 11 (100%) | – | |
WHO stage 4 | 32 (63%) | 26 (65%) | 6 (55%) | 1.54 (0.40–5.98) | |
Previous medication | |||||
Antiretroviral treatment | 14/48 (29%) | 11 (28%) | 3 (27%) | 1.01 (0.22–4.52) | |
TPM/SMZ > 1 month | 8/48 (17%) | 7 (14%) | 1 (9%) | 2.12 (0.23–19.36) | |
Antibiotics | 31 (61%) | 26 (65%) | 5 (45%) | 2.49 (0.64–9.69) | |
Symptom duration, weeks | |||||
< 1 | 12 (24%) | 10 (25%) | 2 (18%) | 1.50 (0.27–8.13) | |
1–3 | 5 (10%) | 4 (10%) | 1 (9%) | 1.11 (0.11–11.08) | |
> 3 | 34 (66%) | 26 (65%) | 8 (73%) | 0.69 (0.15–3.05) | |
P (95% CI)‡ | |||||
Mean age, years | 35 | 34 | 37 | 0.32§ | |
Mean vital signs at first examinations | |||||
O2 saturation (%) | 76 | 75 | 78 | 0.25§ | |
Heart rate/minute | 123 | 123 | 120 | 0.63 (–15.73 to 9.71) | |
Respiratory rate/minute | 46 | 47 | 45 | 0.68 (–10.40 to 6.92) | |
Systolic blood pressure, mm of Hg | 97 | 97 | 96 | 0.45 (–12.64 to 5.69) | |
Mean blood results | |||||
Hemoglobin, g/dL | 9.8 | 10.2 | 8.5 | 0.043 (–3.42 to –0.05) | |
Leukocyte count × 103/μL | 6.7 | 6.9 | 5.8 | 0.52§ | |
Platelet count × 103/μL | 222 | 235 | 171 | 0.18 (–161.45 to 32.63) | |
CD4 cells/μL | 121 | 105 (n = 37) | 204 (n = 7) | 0.66§ |
OR = odds ratio; CI = confidence interval; HIV = human immunodeficiency virus; WHO = World Health Organization; TMP/SMZ = trimethoprim/sulfamethoxazole.
Significant if 95% CI does not include 1.
Significant if 95% CI does not include 0.
Not normally distributed.
Role of bronchoscopy.
Thirty-three patients who were admitted to the MHDU with clinical features of pneumonia were not included in the study because they could not undergo bronchoscopy, either because of early death or because bronchoscopy was contraindicated, usually because of profound hypoxia in the patient (Table 2). Fifty-one patients underwent one bronchoscopy a median of 3 days (range = 1–21 days) after admission to the hospital. Nine (18%) bronchoscopy procedures had to be postponed by 3–5 days because of severe hypoxia (4), wretching/vomiting (2), confusion (1), non-fasted state (1), and unavailability of theatre space (1). Adverse events directly related to the procedure were oxygen desaturation to < 90% in 2 patients (of whom 1 died of confirmed PcP 8 days after bronchoscopy) and a vasovagal reaction in 1 patient.
Results of bronchoscopy and microbiologic investigations of BAL fluid conducted in the context of this study had direct treatment consequences in individual cases when they either confirmed or changed the clinical diagnosis. This finding occurred in 15 (29%) patients who had bacterial pneumonia (n = 4), PCP (n = 2), PTB (n = 3), and PKS (n = 6).
Microbiologic isolates.
Results of microbiologic investigations are shown in Table 4. Eighty-nine positive results from BAL fluid (81) and blood cultures (8) were obtained. Of these results, 15 BAL cultures (13 alpha-haemolytic streptococci, 2 Staphylococcus epidermidis) and 3 blood cultures (2 alpha-haemolytic streptococci, 1 Staphylococcus epidermidis) were considered to be contamination or colonization (not shown in Table 4) because alpha-haemolytic streptococci and Staphylococcus epidermidis are not common causes of pneumonia in patients with advanced HIV infection.20 Additionally, 2 Pneumocystis PCR (Ct > 35) results were also indicative of colonization. Multiple isolates were seen in 15 (29%) patients, 13 of whom had 2 pathogens and 2 patients had 3 pathogens detected.
Table 4.
Patient | Final diagnosis | Confirmed/probable | Co-infection or PKS |
---|---|---|---|
Bacterial pneumonia | |||
1† | Staphylococcus aureus by BAL culture | Confirmed | PcP by PCR |
2 | Klebsiella pneumoniae by BAL culture | Confirmed | PcP by PCR |
3 | Streptococcus pneumoniae by BAL culture | Confirmed | |
4 | Staphylococcus aureus by BAL culture | Confirmed | Parainfluenzavirus type 1 |
5 | Staphylococcus aureus by BAL culture | Confirmed | PKS |
6 | Staphylococcus aureus by BAL culture | Confirmed | PKS |
7 | Klebsiella pneumoniae by BAL culture and Streptococcus pneumoniae by BC | Confirmed | Coronavirus OC 43 |
8 | Staphylococcus aureus by BAL culture | Confirmed | |
9 | Streptococcus pneumoniae by BC | Confirmed | Respiratory syncytial virus A |
10–11 | Streptococcus pneumoniae by BC | Confirmed | |
12–15† (one death) | No isolate | Probable | |
Pneumocystispneumonia | |||
16† | IF/PCR | Confirmed | Pseudomonas by BAL culture |
17 | IF/PCR | Confirmed | Rhinovirus |
18† | IF/PCR | Confirmed | Cryptococcus neoformans by BC + PKS |
19–25 | IF/PCR | Confirmed | |
26† | IF/no PCR | Confirmed | Rhinovirus + PKS |
27 | PCR | Probable | Coronavirus NL |
28 | PCR | Probable | |
29 | PCR | Probable | |
Pulmonary tuberculosis | |||
30† | ZN/mycobacterial culture | Confirmed | MRSA by BAL culture + rhinovirus |
31–33† (two deaths) | ZN/mycobacterial culture | Confirmed | Enterobacter cloacae by BAL culture |
34 | ZN/mycobacterial culture | Confirmed | |
35 | ZN | Confirmed | PKS |
36† | Mycobacterial culture | Confirmed | Enterobacter cloacae by BAL culture |
37 | Mycobacterial culture | Confirmed | PcP by PCR |
38 | Mycobacterial culture | Confirmed | |
39–40 | No isolate | Probable | |
Viral pneumonitis | |||
41 | Influenza A virus | Probable | |
42 | Bocavirus | Probable | |
43† | Adenovirus | Probable | |
Pulmonary Kaposi's sarcoma | |||
44 | Bronchoscopy | Confirmed | Rhinovirus |
45–51† (one death) | Bronchoscopy | Confirmed |
PKS = pulmonary Kaposi's sarcoma; BAL = bronchioalveolar lavage; PcP = Pneumocystis jirovecii pneumonia; PCR = polymerase chain reaction; BC = blood culture; IF = immunofluorescence; ZN = Ziehl-Neelsen staining; MRSA = methicillin-resistant S. aureus; all viral isolates by PCR.
Died.
Final diagnoses.
The final clinical diagnoses in 51 patients were bacterial pneumonia, n = 15 (29%); PcP, n = 14 (27%); PTB, n = 11 (22%); PKS, n = 8 (16%), and viral pneumonitis, n = 3 (6%). In 39 (77%) of the cases, the diagnosis could be confirmed (Table 4). Diagnoses of viral pneumonitis and two cases of smear-negative but mycobacterial-culture positive PTB were made retrospectively because these test results were not available in real-time. Diagnoses for the three HIV-negative persons were confirmed pneumococcal pneumonias in two and probable influenza A viral pneumonitis in the third patient. All PCRs in BAL fluid for the atypical agents Chlamydophila pneumoniae, Chlamydophila psittaci, Mycoplasma pneumoniae, and Legionella pneumophila were negative.
Clinical features and outcome.
Clinical features and outcome of all 51 final diagnoses are shown in Table 5. Hemoptysis and pleural effusion were not seen in any patients with PcP, and patients with PcP differed from those with other diagnoses in their mean hemoglobin concentration (11.8 g/dL versus 9.1 g/dL; P < 0.0001) and mean CD4 count (17 versus 169 cells/μL; P < 0.0001). None of the 14 patients with PcP was receiving ART, and only one had been taking trimethoprim/sulfamethoxazole prophylaxis, which he had stopped two weeks before admission. In contrast, most patients with PKS had hemoptysis and pleural effusion and 6 of 8 had been established on ART. In-hospital mortality was similar for all diagnoses. Only one (8%) of 13 patients with PKS had no dermal or oral KS lesions.
Table 5.
Characteristic | Bacterial pneumonia (n = 15) | Pneumocystis pneumonia (n = 14) | Pulmonary tuberculosis (n = 11) | Pulmonary Kaposi's sarcoma (n = 8) |
---|---|---|---|---|
Female sex | 7 (47%) | 8 (57%) | 5 (45%) | 2 (25%) |
HIV status known at study entry | 7 (47%) | 7 (50%) | 9 (82%) | 7 (88%) |
Confirmed HIV positive | 13 (87%) | 4 (100%) | 9 (100%) | 8 (100%) |
ART at study entry | 3 (20%) | 0; P = 0.005† | 5 (45%) | 6 (75%; P = 0.003)† |
Hemoptysis | 4 (27%) | 0; P = 0.022† | 3 (27%) | 4 (50%) |
Crackles on chest auscultation | 13 (87%) | 7 (50%); P = 0.005)† | 9 (82%) | 8 (100%) |
Pleural effusion | 6 (40%) | 0; P = 0.011† | 2 (18%) | 7 (88%); P = 0.019† |
Survived | 13 (87%) | 11 (79%) | 7 (64%) | 7 (88%) |
Mean | ||||
Oxygen saturation %‡ | 71 | 70 | 81 | 85; P = 0.017† |
Heart rate/minute | 122 | 132; P = 0.04† | 119 | 116 |
Respiratory rate/minute | 51; P = 0.046† | 46 | 47 | 42 |
Systolic blood pressure, mm of Hg | 93 | 94 | 98 | 104 |
Hemoglobin, g/dL | 9.6 | 11.8; P < 0.0001† | 8.0; P = 0.03† | 8.6 |
Leukocyte count × 103 μL‡ | 8.8 | 6.1 | 5.3 | 6.2 |
Platelet count × 103 μL | 237 | 291; P = 0.034† | 171 | 115 |
Lymphocyte count × 103 μL‡ | 1.7 | 0.8 | 0.7; P = 0.02† | 1.5 |
CD4 cells/μL‡ | 168; P = 0.027† | 17; P < 0.0001† | 157 | 201 |
CD4 cells/μL‡ | ||||
< 100 | 5 | 14 | 6 | 5 |
100–199 | 2 | 1 | ||
> 200 | 4 | 3 | 2 |
Each diagnosis was compared with a composite of the other four diagnoses (3 cases of viral pneumonia are not shown). HIV = human immunodeficiency virus; ART = antiretroviral therapy.
Only P < 0.05 is shown.
Not normally distributed.
Discussion
We comprehensively described microbiologic etiology and clinical features of adult Malawians who were hospitalized with clinical features of pneumonia severe enough to require admission to a MHDU. We confined the study to patients who were judged by traditional criteria to be able to tolerate bronchoscopy and bronchoalveolar lavage, and in these patients the procedure caused no major adverse events and contributed importantly to obtaining a final diagnosis. We were able to confirm a final diagnosis by detection of an infectious pathogen or PKS in a high proportion of the study population (n = 39, 77%), which is comparable to percentages of 59–90% reported in some studies of pneumonia etiology in similar African settings and in the developed world.1,5,6,21 Four main diagnoses of bacterial pneumonia (29%), PcP (27%), PTB (22%), and PKS (16%) in all 51 patients emerged. In one-third of patients multiple pulmonary morbidities were detected exceeding percentages of 10–13% reported from Kenya, Uganda and Tanzania.1,5,6
Despite HIV/AIDS public awareness campaigns and the ongoing ART scale-up in Malawi, a high number of patients (40%) claimed to be unaware of their HIV status at presentation. Twenty-nine percent had started ART, but only 16% were receiving trimethoprim-sulphamethoxazole prophylaxis, an intervention shown to reduce HIV-related mortality.22 At the time of the study, the national trimethoprim/sulfamethoxazole prophylaxis program had not been widely implemented.
The high PcP prevalence among HIV-positive patients hospitalized for severe pneumonia has not been previously documented in Malawi. Two other studies have determined PcP rates in adults in Malawi in different settings. Hargreaves et al. identified PcP by indirect immunofluorescence or PCR of BAL fluid in 9% of 186 AFB smear–negative outpatients who were about to start anti-tuberculous treatment.14 In a prospective community-based cohort study of HIV-positive persons, we previously found an overall incidence rate of 1.0 per 100 person-years observation, and based diagnoses on immunofluorescent staining and Pneumocystis PCR of induced sputum specimens.13 In the subgroup of persons with low CD4 counts (< 100/mm3) PcP was more common (5.0/100 person-years observation). In the present study, one-fourth of our severely ill patients had confirmed or probable Pneumocystis infection, a percentage similar to the 27–38% reported in cross-sectional studies of various designs among AFB smear–negative persons from Uganda, Kenya, Ethiopia, and Zimbabwe.5,7,23,24 These findings confirm that PcP is an important diagnosis in hospitalized HIV-infected patients with severe respiratory illness in sub-Saharan Africa.
Well-documented clinical features of PcP including hypoxemia, unremarkable chest auscultation, and absence of hemoptysis and pleural effusion were also seen in the present study.25 The most helpful laboratory indicator other than BAL IF for the presence of PcP was a CD4 count < 100 cells/mm3. Of the 16 clinically relevant positive PcP IF and/or PCR results for which a CD4 count was available, 14 (88%) were measured as < 25/mm3 whereas this was observed in only 2 (7%) of 28 patients who had no evidence of PcP infection, which supports the routine use of CD4 counts when evaluating HIV-infected patients with suspected severe pneumonia in settings in sub-Saharan Africa. Most patients had begun high-dose trimethoprim/sulfamethoxazole treatment before the positive result of the real-time IF stain became available, indicating a high index of clinical suspicion among clinicians. Inpatient mortality of patients with PcP was 21% (n = 3) which is comparable to mortality figures reported by others and similar to patients in our study with diagnoses other than PcP.7,24 On-site and real-time IF methods for confirmation of PcP proved to be a useful diagnostic tool, especially because half of the patients with PcP had a secondary infection or PKS. Three patients who already had another primary diagnosis were also treated with trimethoprim/sulfamethoxazole for PcP on clinical grounds, and proved on examination of BAL fluid to have a negative IF test result but a positive PCR result for Pneumocystis (Ct < 35). The interpretation of these findings is uncertain because colonization of airways with Pneumocystis is well-described in immunocompromised persons.26
Sputum smear AFB-negative PTB remains a challenging entity because no validated diagnostic or management algorithms exist.27 Almost one-fifth of our sputum smear-negative patients had microbiologic proof of tuberculosis upon further investigation, which is comparable to previous reports citing detection rates of up to 33% on BAL or induced sputum specimens.6,23,28 We may have under-diagnosed PTB by sputum smear examination because we only examined one expectorated sputum sample before performing bronchoscopy. Other factors contributing to the low sensitivity of sputum smear examination of approximately 20–30% include advanced immunosuppression, inability to expectorate, and variable quality of sampling and specimen processing procedures.29 Mycobacterial culture increases diagnostic sensitivity, but it is rarely available in sub-Saharan Africa. Of five BAL AFB smear-negative patients in our study, three were culture positive, two of whom had not received anti-tuberculous treatment by the time the culture results became known. Non-specific and atypical clinical features in patients who recently began treatment with ART and subsequently having IRIS can complicate the diagnosis of sputum smear-negative PTB. Five of 11 patients with the final diagnosis of PTB were within two months of having started treatment with ART previously, and IRIS could have contributed to their current illness. Profound anemia was a distinguishing clinical marker separating PTB from other diagnoses, a finding that has been reported elsewhere.6
Most (84%) patients at admission were treated empirically for suspected bacterial pneumonia. However, only one-third of these cases had a positive BAL or blood culture result. The predominant pyogenic isolates from BAL fluid were Staphylococcus aureus (n = 6), Pseudomonas aerugenosa (n = 1), Enterobacter cloacae (n = 2), and Klebsiella pneumonia (n = 2). A higher prevalence of community-acquired atypical pyogenic bacteria in patients with advanced HIV infection has been reported by other investigators.5,6,20 Some of the isolates in our study might have been hospital acquired because for 30% of the patients, BAL sampling was carried out after five days of hospitalization. Streptococcus pneumoniae was identified in four blood cultures, which had been obtained at admission, whereas only one BAL sample was positive. We may have under-diagnosed pneumococcal disease on BAL fluid specimens in our population because two-thirds of patients had received antibiotic treatment before hospital admission and further empirical antibiotic therapy was commonly commenced in hospital before bronchoscopy was undertaken. Given the unconventional bacterial spectrum observed in this and other studies, antimicrobial treatment with a narrow-spectrum penicillin for severe community acquire pneumonia in HIV-positive patients might not be sufficient.
Viral causes of community-acquired pneumonia have received heightened attention in the western world and viral detection rates of up to 29% in HIV-negative patients have been reported.30 Hardly any epidemiologic data have originated from sub-Saharan Africa where sensitive PCR-based diagnostics are seldom used. Our study is the first to report the results of comprehensive PCR assays for detection of respiratory viruses. Scott et al. retrospectively identified infection with influenza A and B and adenovirus by serologic testing in 5.7% of adult patients in Kenya with acute pneumonia, half of whom had bacterial co-infections.1 The presence of viral nucleic acid in BAL or upper airway swabs is generally believed to represent infection, although asymptomatic carrier states have been described. The viruses detected in our cohort have been associated with respiratory illness to a variable extent, although it remains controversial whether viruses such as rhinovirus, coronaviruses, or the newly described bocavirus cause pneumonia through invasion and replication in the lower respiratory tract, facilitate bacterial infection or merely are innocent bystanders.30–33 We documented two deaths in four patients who had evidence of rhinovirus infection, although both had significant co-infections with PcP and PTB/methicillin-resistant Staphylococcus aureus. Both patients infected with bocavirus (no other organism isolated) and coronavirus NL (co-infection with PcP) infection survived to discharge.
There is also a paucity of data from Africa concerning atypical bacterial organisms as a cause of pneumonia. Lockman et al. used PCR testing and convalescence serologic analysis to identify 36 (17%) cases of acute Mycoplasma pneumonia among hospitalized, predominantly HIV-infected adults in urban Botswana, of whom more than two-thirds had a co-pathogen detected.8 A much lower yield of 2.5%, diagnosed by only by using convalescence serologic analysis, was seen in a cohort of patients with radiologically confirmed acute community-acquired pneumonia in Kenya, 52% of whom were HIV-infected.1 Neither study found evidence of infection with Chlamydophila pneumoniae. Somewhat unexpectedly, we were unable to identify any Mycoplasma pneumoniae, Chlamydophila pneumoniae, Legionella pneumophila, or Chlamydophila psittaci by PCR testing. Although DNA PCR as the sole diagnostic test, as used in our study, may have a lower sensitivity between 78% and 90% compared with serologic analysis, it seems unlikely that an important number of cases in our study were missed.8 It may be that the data collection occurred outside a Mycoplasma epidemic or that the generally less severe clinical features associated with Mycoplasma disease obviated hospital and particularly MHDU admission. However, it is reasonable to assume that atypical bacteria are unlikely to represent a common cause of severe respiratory infections in Malawi.
Intrathoracic KS occurs in up to 10% of patients with AIDS and this increases to 25% if mucocutaneous involvement is present.34 The incidence of KS has decreased in Europe and North America since the advent of ART, a trend that could not have been apparent in Malawi at the time of this study, when ART scale-up was not yet advanced. One-fourth of our patients displayed characteristic features of tracheobronchial KS, all of whom had markedly abnormal chest radiographs. Endobronchial KS (or pulmonary KS once lung parenchyma has been infiltrated) is believed to develop after mucocutaneous KS is established, a fact that can aid the diagnosis in patients with respiratory symptoms in which bronchoscopy for inspection of large airways is not available. All 12 patients in our study with dermal and/or oral mucosal KS (both in 8 patients) also had tracheobronchial disease, and only one patient had endobronchial KS without skin or oral involvement. Therefore, the triad of severe respiratory symptoms, abnormal chest radiograph, and presence of mucocutaneous KS was strongly associated with pulmonary KS in this patient cohort. However, intrathoracic KS associated respiratory symptoms, radiographic infiltrates and/or pleural effusion (present in 53% of all 13 patients with evidence of KS in our study) are non-specific findings and extensive investigations to exclude opportunistic pulmonary infections are recommended, as highlighted by the fact that 6 of the 13 patients with PKS in our study had co-diagnoses of PcP, Staphylococcus aureus pneumonia, PTB, or rhinovirus infection.35 In five PKS patients respiratory illness developed soon after ART initiation and without co-infection, suggesting the possibility of IRIS.36
Our study had several limitations. First, we investigated a select group of severely ill and hypoxic, sputum AFB smear-negative patients who came to an urban tertiary referral center. Therefore, extrapolation of our findings to other settings is not straightforward, although many patient characteristics were comparable to studies carried out elsewhere in sub-Saharan Africa.5,6 Second, in the absence of validated tools to assess severity of community acquired pneumonia in sub-Saharan Africa such as CURB 65, which is used in the western world, direct comparison of patient cohorts from different studies is fraught with difficulty.37 Third, pre-admission and in-hospital empirical antibiotic treatment before carrying out study investigations will have influenced the detection rate and spectrum of organisms we found. Fourth, almost one-third of patients had commenced ART recently, yet we were unable to assess formally whether the illness occurred in the context of IRIS. Fifth, because the number of patients described was small, any statistical significance of results pertaining to even smaller subgroups should be viewed with caution. And finally, we did not have an independent review of case notes, radiologic findings, and final diagnoses. Interpretation of results is therefore based on the expert opinion of the authors.
In a time of ongoing ART scale-up, we comprehensively determined the etiology of severe pneumonia in mostly HIV-positive adults in Malawi at a high dependency unit of an urban tertiary hospital. Investigations including bronchoscopy and BAL could be safely performed and provided a confirmed final diagnosis in three-fourths of the patients. Most diagnoses were PcP, PTB, bacterial pneumonia, and PKS, and these occurred at similar frequencies. Mycoplasma, Chlamydophila, and Legionella species did not play a role in this patient group. The clinical relevance of observed viral isolates is not entirely clear and requires further study.
ACKNOWLEDGMENTS
We thank the patients, nursing staff, and endoscopy suite personnel for support and participation. The American Committee on Clinical Tropical Medicine and Travellers' Health (ACCTMTH) assisted with publication expenses.
Footnotes
Authors' addresses: Thomas K. Hartung, Institute of Tropical Medicine and International Health, Charité – Universitätsmedizin, Berlin, Germany, E-mail: thomas.hartung@nhs.net. Daniel Chimbayo, Lifeline Malawe, PO Box 41, Chipoka, Salima, Malawi, E-mail: danielchimbayo@yahoo.com. Joep J. G. van Oosterhout and Ed E. Zijlstra, Department of Medicine, College of Medicine, University of Malawi, Blantyre, Malawi, E-mail: joepvanoosterhout@gmail.com and e.e.zijlstra@gmail.com. Tarsizio Chikaonda and Malcom E. Molyneux, Malawi-Liverpool Wellcome Trust Clinical Research Program, PO Box 30096, Chichiri, Blantyre 3, Malawi, E-mails: t.chikaonda@yahoo.co.uk and mmolyneux999@gmail.com. Gerard J. J. van Doornum, Department of Virology, Unit Diagnostics, L359, Erasmus Medical Centre, Gravendijkwal 230, 3015 CE Rotterdam, The Netherlands, E-mail: g.vandoornum@erasmusmc.nl. Eric C. J. Claas, Department of Medical Microbiology, Leiden University Medical Centre, Albinusdreef 2, 2333 ZA Leiden, The Netherlands, E-mail: e.c.j.claas@lumc.nl. Willem J. G. Melchers, Department of Medical Microbiology, Radboud University Nijmegen Medical Centre, PO Box 9101, 6500HB Nijmegen, The Netherlands, E-mail: w.melchers@mmb.umcn.nl.
References
- 1.Scott JA, Hall AJ, Muyodi C, Lowe B, Ross M, Chohan B, Mandaliya K, Getambu E, Gleeson E, Gleeson F, Drobniewski F, Marsh K. Aetiology, outcome and risk factors for mortality among adults with acute pneumonia in Kenya. Lancet. 2000;355:1125–1130. doi: 10.1016/s0140-6736(00)02089-4. [DOI] [PubMed] [Google Scholar]
- 2.Murray JF. Pulmonary complications of HIV-infection among adults living in sub-Saharan Africa. Int J Tuberc Lung Dis. 2005;9:826–835. [PubMed] [Google Scholar]
- 3.Shelburne SA, Visnegarwala F, Darcourt J, Graviss EA, Giordano TP, White AC, Jr, Hamill RJ. Incidence and risk factors for immune reconstitution inflammatory syndrome during highly active antiretroviral therapy. AIDS. 2005;19:399–406. doi: 10.1097/01.aids.0000161769.06158.8a. [DOI] [PubMed] [Google Scholar]
- 4.Bonnet MM, Pinoges LL, Varaine FF, Oberhauser BB, O'Brien DD, Kebede YY, Hewison CC, Zachariah RR, Ferradini LL. Tuberculosis after HAART initiation in HIV-positive patients from five countries with a high tuberculosis burden. AIDS. 2006;20:1275–1279. doi: 10.1097/01.aids.0000232235.26630.ee. [DOI] [PubMed] [Google Scholar]
- 5.Worodria W, Okot-Nwang M, Yoo SD, Aisu T. Causes of lower respiratory infection in HIV-infected Ugandan adults who are sputum AFB smear-negative. Int J Tuberc Lung Dis. 2003;7:117–123. [PubMed] [Google Scholar]
- 6.Kibiki GS, Berckers P, Mulder T, Arens T, Mueller A, Boeree MJ, Shao JF, Van Der Ven AJ, Diefendahl H, Dolmans WM. 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. 2007;84:420–428. doi: 10.4314/eamj.v84i9.9551. [DOI] [PubMed] [Google Scholar]
- 7.Chakaya JM, Bii C, Ng'ang'a L, Amukoye E, Ouko T, Muita L, Gathua S, Gitau J, Odongo I, Kabanga JM, Nagai K, Suzumura S, Sugiura Y. Pneumocystis carinii pneumonia in HIV/AIDS at an urban district hospital in Kenya. East Afr Med. 2003;80:30–35. doi: 10.4314/eamj.v80i1.8663. [DOI] [PubMed] [Google Scholar]
- 8.Lockman S, Hone N, Kenyon TA, Mwasekaga M, Villauthapillai M, Creek T, Zell E, Kirby A, Thaker WL, Talkington D, Moura IN, Binkin NJ, Clay L, Tappero JW. Etiology of pulmonary infections in predominantly HIV-infected adults with suspected tuberculosis, Botswana. Int J Tuberc Lung Dis. 2003;7:714–723. [PubMed] [Google Scholar]
- 9.Crampin AC, Glynn JR, Fine PEM. What has Karonga taught us? Tuberculosis studied over three decades. Int J Tuberc Lung Dis. 2009;13:153–164. [PMC free article] [PubMed] [Google Scholar]
- 10.Gordon S, Graham S. Epidemiology of respiratory disease in Malawi. Malawi Medical Journal. 2006;18:134–146. doi: 10.4314/mmj.v18i3.10918. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Gordon SB, Chaponda M, Walsh AL, Whitty CJM, Gordon M, Machili CE, Gilks CF, Boeree MJ, Kampondeni S, Read RC, Molyneux ME. Pneumococcal disease in HIV-infected Malawian adults: acute mortality and long-term survival. AIDS. 2002;16:1409–1417. doi: 10.1097/00002030-200207050-00013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.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. doi: 10.1016/S0140-6736(98)11074-7. [DOI] [PubMed] [Google Scholar]
- 13.van Oosterhout JJ, Laufer ML, Thesing PC, Perez MA, Graham SM, Chimbiya N, Thesing PC, Alvarez-Martinez MJ, Wilson PE, Zijlstra EE, Taylor TE, Plowe CV, Meshnick SR. Pneumocystis pneumonia in HIV-positive adults, Malawi. Emerg Infect Dis. 2007;13:325–328. doi: 10.3201/eid1302.060462. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Hargreaves NJ, Kadzakumanja O, Phiri S, Lee CH, Tang X, Salaniponi FM, Harries AD, Squire SB. Pneumocystis carinii pneumonia in patients being registered for smear-negative pulmonary tuberculosis in Malawi. Trans R Soc Trop Med Hyg. 2001;95:402–408. doi: 10.1016/s0035-9203(01)90197-x. [DOI] [PubMed] [Google Scholar]
- 15.Lewis DK, Callaghan M, Phiri K, Chipwete J, Kublin JG, Borgstein E, Zijlstra EE. Prevalence and indicators of HIV and AIDS among adults admitted to medical and surgical wards in Blantyre, Malawi. Trans R Soc Med Hyg. 2003;97:91–96. doi: 10.1016/s0035-9203(03)90035-6. [DOI] [PubMed] [Google Scholar]
- 16.Honeybourne D, Babb J, Bowie P, Brewin A, Fraise A, Garrard C, Harvey J, Lewis R, Neumann C, Wathen CG, Williams T. British Thoracic Society guidelines on diagnostic flexible brochoscopy. Thorax. 2001;56:i1–i21. doi: 10.1136/thorax.56.suppl_1.i1. The British Thoracic Society Guidelines Committee. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Bosis S, Esposito S, Niesters HG, Zuccotti GV, Marseglia G, Lanari M, Zuin G, Pelucchi C, Osterhaus AD, Principi N. Role of respiratory pathogens in infants hospitalized for a first episode of wheezing and their impact on recurrences. Clin Microbiol Infect. 2008;14:677–684. doi: 10.1111/j.1469-0691.2008.02016.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Linssen CF, Jacob JA, Beckers P, Templeton KE, Bakkers J, Kuijper EJ, Melchers WJ, Drent M, Vink C. Inter-laboratory comparison of three different real-time PCR assays for the detection of Pneumocystis jirovecii in bronchoalveolar lavage fluid samples. J Med Microbiol. 2006;55:1229–1235. doi: 10.1099/jmm.0.46552-0. [DOI] [PubMed] [Google Scholar]
- 19.Flori P, Bellete B, Durand F, Raberin H, Cazorla C, Hafid J, Lucht F, Tran Manh Sung R. Comparison between real-time PCR, conventional PCR and different staining techniques for diagnosing Pneumocystis jirovecii pneumonia from bronchoalveolar specimens. J Med Microbiol. 2004;53:603–607. doi: 10.1099/jmm.0.45528-0. [DOI] [PubMed] [Google Scholar]
- 20.Maynaud C, Parrot A, Cadranel J. Pyogenic bacterial lower respiratory tract infection in human immunodeficiency virus-infected patients. Eur Respir J. 2002;20:28S–39S. doi: 10.1183/09031936.02.00400602. [DOI] [PubMed] [Google Scholar]
- 21.Lim WS, Baudouin SV, George RC, Hill AT, Jamieson C, Le Jeune I, Macfarlane JT, Read RC, Roberts HJ, Levy ML, Wani M, Woodhead MA. British Thoracic Society guidelines for the management of community acquired pneumonia in adults: update 2009. Thorax. 2009;64:iii1–iii55. doi: 10.1136/thx.2009.121434. [DOI] [PubMed] [Google Scholar]
- 22.Anglaret X, Chene G, Attia A, Toure S, Lafont S, Combe P, Manlan K, N'Dri-Yoman T, Salamon R. Early chemoprophylaxis with trimethoprim-sulphamethoxazole for HIV-1-infected adults in Abidjan, Côte d'Ivoire: a randomized trial. Cotrimo-CI Study Group. Lancet. 1999;353:1463–1468. doi: 10.1016/s0140-6736(98)07399-1. [DOI] [PubMed] [Google Scholar]
- 23.Aderaye G, Bruchfeld J, Aseffa G, Nigussie Y, Melaku K, Woldeamanuel Y, Asrat D, Worku A, Gaegziabher H, Lebaad M, Lindquist L. Pneumocystis jirovecii pneumonia and other pulmonary infections in TB smear negative HIV positive patients with atypical chest X-ray in Ethiopia. Scand J Infect Dis. 2007;39:1045–1053. doi: 10.1080/00365540701474508. [DOI] [PubMed] [Google Scholar]
- 24.Malin AS, Gwanzura LK, Klein S, Robertson VJ, Musvaire P, Mason PR. Pneumocystis carinii pneumonia in Zimbabwe. Lancet. 1995;346:1258–1261. doi: 10.1016/s0140-6736(95)91862-0. [DOI] [PubMed] [Google Scholar]
- 25.Fisk DT, Meshnick S, Kazanjian PH. Pneumocystis carinii pneumonia in patients in the developing world who have acquired immunodeficiency syndrome. CID. 2003;36:70–78. doi: 10.1086/344951. [DOI] [PubMed] [Google Scholar]
- 26.Morris A, Lundgren JD, Masur H, Walzer PD, Hanson DL, Frederick T, Huang L, Beard CB, Kaplan JE. Current epidemiology of Pneumocystis pneumonia. Emerg Infect Dis. 2004;10:1713–1720. doi: 10.3201/eid1010.030985. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Colebunders R, Bastian I. A review of the diagnosis and treatment of smear-negative pulmonary tuberculosis. Int J Tuberc Lung Dis. 2000;4:97–107. [PubMed] [Google Scholar]
- 28.Hartung TK, Maulu A, Nash J, Fredlund VG. Suspected pulmonary tuberculosis in rural South Africa. Sputum induction as a simple diagnostic tool? S Afr Med J. 2002;92:455–458. [PubMed] [Google Scholar]
- 29.Harries AD. Tuberculosis in Africa: clinical presentation and management. Pharmacol Ther. 1997;73:1–50. doi: 10.1016/s0163-7258(96)00115-5. [DOI] [PubMed] [Google Scholar]
- 30.Jennings LC, Anderson TP, Beynon KA, Chua A, Laing RTR, Wemo AM, Young SA, Chambers ST, Murdoch DR. Incidence and characteristics of viral community-acquired pneumonia in adults. Thorax. 2008;63:42–48. doi: 10.1136/thx.2006.075077. [DOI] [PubMed] [Google Scholar]
- 31.Marcos MA, Esperatti M, Torres A. Viral pneumonia. Curr Opin Infect Dis. 2009;22:143–147. doi: 10.1097/QCO.0b013e328328cf65. [DOI] [PubMed] [Google Scholar]
- 32.Longtin J, Bastien M, Gilca R, Leblanc E, de Serres G, Bergeron MG, Boivin G. Human bocavirus infection in hospitalized children and adults. Emerg Infect Dis. 2008;14:217–221. doi: 10.3201/eid1402.070851. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.van der Hoek L. Human coronaviruses: what do they cause? Antivir Ther. 2007;12:651–658. [PubMed] [Google Scholar]
- 34.Aboulafia DM. The epidemiologic, pathologic and clinical features of AIDS-associated pulmonary Kaposi's sarcoma. Chest. 2000;117:1128–1145. doi: 10.1378/chest.117.4.1128. [DOI] [PubMed] [Google Scholar]
- 35.Miller RF, Tomlinson MC, Cottrill CP, Donals JJ, Spittle MF, Semple SJ. Bronchopulmonary Kaposi's sarcoma in patients with AIDS. Thorax. 1992;47:721–725. doi: 10.1136/thx.47.9.721. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Muller M, Wandel S, Colebunders R, Attia S, Furrer H, Eggert M. Immune reconstitution inflammatory syndrome in patients starting antiretroviral therapy for HIV infection: a systematic review and meta-analysis. Lancet Infect Dis. 2010;10:251–261. doi: 10.1016/S1473-3099(10)70026-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Lim WS, van der Eerden MM, Laing R, Boersma WG, Karalus N, Town GI, Lewis SA, MacFarlane JT. Defining community acquired pneumonia severity on presentation to hospital: an international derivation and validation study. Thorax. 2003;58:377–382. doi: 10.1136/thorax.58.5.377. [DOI] [PMC free article] [PubMed] [Google Scholar]