Professional Documents
Culture Documents
higher (7). The prognosis of HIV children with PCP monary symptoms (i.e. cough or dyspnea), (ii) pulmo-
depends on multiple factors, including degree of nary infiltration observed by chest radiography or
hypoxemia at presentation, need for mechanical venti- computed tomography, (iii) detection of P. jiroveci cyst
lation and access to medical care (4). However, to the in the BAL samples, (iv) receipt of antimicrobial
best of our knowledge, few studies have recorded and therapy for PJP during hospitalization, (v) negative
analyzed the clinical characteristics and outcome of HIV tests, (vi) only children with primary infections
PCP in non-HIV children. were included. The study excluded children with
Early appropriate prophylaxis for PCP in non-HIV incomplete medical information.
patients is crucial. Several published studies have
revealed that the most significant risk factor for PCP in Data collection
non-HIV patients is the use of glucocorticoid therapy Clinical data collected included the following: general
(8–10). However, practices of PCP prophylaxis in chil- demographic information, underlying diseases, immu-
dren are mostly extrapolated from studies that have nosuppressive therapies, clinical symptomatology, lab-
used adult subjects. The appropriate time-point to oratory values and hospital mortality.
start prophylaxis treatment is still being studied, espe- Demographical and clinical details were retrieved
cially in non-HIV children receiving glucocorticoid. from the hospital’s computerized record system. For
Although a variety of serum markers have been stud- each patient, clinical symptoms and time from clinical
ied for evaluation of PCP in non-HIV adult patients, onset to diagnosis were recorded.
few serum markers have been examined in non-HIV High dose glucocorticoid was defined as 2 mg/kg/
children (5, 11). Data suggest that low peripheral day for at least 14 days or methylprednisolone pulse
blood CD4 counts may help identify non-HIV patients therapy for 3–5 days.
at risk for PCP. However, CD4 counts may be inaccu- CD4%, CD8% and CD4/CD8 were recorded and the
rate in neutrophilic states (5, 10, 12, 13). Some lowest counts in the two weeks preceding diagnosis
research reported that CD4/CD8 ratio significantly were used for analysis.
decreased when PCP developed (14, 15). We specu- PCP was considered the cause of mortality when the
lated that CD4/CD8 ratio may be a serum marker for treating physician recorded acute respiratory distress
PCP in non-HIV children. syndrome (ARDS) or ARF (acute respiratory failure)
The aims of this retrospective study were to (i) as cause of death after PCP was diagnosed within one
observe the clinical of characteristics and outcome of month.
PCP in non-HIV children, (ii) assess the utility of In order to identify factors affecting PCP outcome,
CD4/CD8 as a marker for initiation of PCP prophy- laboratory data of survivors were compared with non-
laxis in non-HIV children and (iii) identify clinical fac- survivors.
tors associated with PCP mortality.
Co-infections
Patients and methods
If children were on the anti-tuberculosis treatment or
Study population tested strong positive in the purified protein derivative
We retrospectively evaluated data from children (under (PPD) test, we considered it as tuberculosis infection.
15 years old) who were diagnosed with PCP at Beijing If children have thrush, we considered it as Candida
Children’s Hospital, during the period from 1 January albicans infection.
2005 until 31 December 2014. The diagnosis of PCP Microbiological investigations performed on BAL
was made via immunofluorescence or Giemsa staining included Gram’s stain, acid-fast stain and cultures for
of P. jirovecii cysts in BAL (bronchoalveolar lavage conventional bacteria, mycobacteria and fungi. Quan-
fluid). BAL fluids were obtained using fiberoptic bron- titative culture of BAL specimens was used for assaying
choscopy as follows: after examination of tracheobron- conventional bacteria, and the threshold value was
chial tree, normal saline (20–50 mL) was injected and 104 colony-forming units (CFU)/mL.
20–50 mL aliquots were manually aspirated, usually
from middle lobe or lingula airways. PCP-PCR was Control group
not used as a diagnostic tool. We selected 120 children who underwent BAL for the
diagnosis of PCP during the study period as control
Inclusion criteria
subjects (Table 1). All of them had the presence of pul-
Six inclusion criteria had to be met by each child monary symptoms (i.e. cough or dyspnea) and pulmo-
included in this study: (i) the presence of relevant pul- nary infiltration observed by chest radiography or
computed tomography. However, BAL results were Windows statistical software package (SPSS Inc., Chi-
negative for PCP. In order to assess the value of CD4/ cago, IL).
CD8 to identify PCP from other pneumonia, we
recorded CD4 count, CD4%, CD8% and CD4/CD8 Result
from all subjects with other pneumopathy (including Part 1: Clinical characteristics of the PCP
bacterial pneumonia, tuberculosis pneumonia, fungal pediatric patients
pneumonia, viral pneumonia).
From January 2005 to December 2014, we identified
69 PCP-infected children admitted to our hospital.
Statistical analysis None of them were receiving prophylaxis at the time of
diagnosis. Four of them were infected with HIV, thus
The variables in the dataset were described or summar- excluded from the study. Five subjects were excluded
ized by using either median and interquartile range or on the basis of incomplete medical history data. We
number and proportion of the total (%). To determine recorded demographics, clinical characteristics and
the association of independent variables with hospital underlying diseases of the remaining 60 non-HIV
mortality, continuous variables were compared using PCP-infected pediatric patients (Tables 1 and 2).
the Student’s two-tailed t test or non-parametric A total of 45 (75.0%) of the subjects were receiving
Mann–Whitney U-test, in case of non-normal distri- immunosuppressants for their underlying diseases.
bution. Chi-square test or Fisher’s exact test, in case of Glucocorticoid alone was administered in 16 children
low expected frequencies, was used for comparisons of (35.6%), chemotherapeutic agents alone were adminis-
categorical variables. We identified variables as signifi- tered in 5 children (11.1%) and glucocorticoids com-
cant in the univariate analysis, and they were assessed bined with immunosuppressive or chemotherapeutic
as predictors of mortality using logistic regression agents were administered in 24 children (53.3%). PCP
analysis. To evaluate the sensitivity and specificity occurred in 23/40 (57.5%) children after they received
of serum marker, receiver operating characteristic an additional high dose of glucocorticoid because of
(ROC) curves were constructed for CD4/CD8. The recurrent disease; 3/40 (7.5%) when receiving initiated
areas under the ROC curves were compared in a glucocorticoid treatment; 8/40 (20%) during cortico-
non-parametric approach. All statistical tests were two- steroid tapering; 6/40 (15%) during low dose mainte-
sided, and P < 0.05 as considered statistically signifi- nance of glucocorticoids. The cumulative dose of
cant. All data were analyzed with SPSS, version 17 for corticosteroid when PCP developed was 239.98 mg/kg
Table 2. Underlying diseases at diagnosis of Pneumocystis tively) in all 60 PCP-infected children compared with
pneumonia 120 non-PCP children, while CD8% was increased
Number of (44.38 6 12.93 vs 27.70 6 13.27, P < 0.05).
Disease patients (%) ROC curves were used to assess the potential utility
Connective tissue disease 23 (38.3) of CD4/CD8 ratio detection in children with PCP. The
Juvenile idiopathic arthritis 7 area under the ROC curve of CD4/CD8 levels for the
Systemic lupus erythematosus 5 diagnosis of PCP was 0.902 (95% CI, 0.849–0.955).
Juvenile dermatomyositis 5 The analysis rendered an optimum cut-off value of
Takayasu’s Arteritis 4 0.715 corresponding to 89.2% sensitivity and 80.4%
Wegener’s granulomatosis 2
specificity (Fig. 1).
Hematological diseases 14 (23.3)
Acute lymphoid leukemia 7
Immune thrombocytopenic purpura 3
Part 3. PCP treatment and outcome
Acute myeoodi leukemia 1 All the PCP-infected children (n 5 60) were treated with
Aplastic anemia 1 TMP/SMZ when PCP was confirmed. In order to define
Hemophagocytic syndrome 1
prognosis factors of PCP, 25 children who died were
Allogeneic haematopoietic stem 1
cell transplantation
compared with 35 children who survived (Table 3).
Nephrotic disease 8 (13.3) Age; sex; underlying immune defect and immuno-
Nephrotic syndrome 6 suppressive drug; the duration between the onset of
IgA nephropathy 2 symptoms and initiation of appropriate treatment;
Immunodeficiency disease 10 (16.7) ALB, Hb, CRP, CD4 counts; and admittance to
X-linked hyperimmunoglobulin M syndrome 6 PICU were not associated with a poor prognosis.
severe combined immunodeficiency 4 Using a multivariate logistic regression model (Table
Other 5 (8.3)
4), the LDH level, need for mechanical ventilation
Dermatologic disease 3
Diabetes mellitus 1
and co-infection were associated with a poor prog-
Surgical operation 1 nosis of PCP in non-HIV children.
Discussion
(range, 28–1890 mg/kg). The median time from begin-
ning immunosuppressive medication to PCP diagnosis To the best of our knowledge, this is the first study that
was 245.9 days (range: 14–2100 days). examines PCP in non-HIV children. This retrospective
Laboratory data including lactic dehydrogenase
(LDH), C-reactive protein (CRP), neutrophile percent
(neu%), CD4 counts, CD4%, CD8% and CD4/CD8
ratio were available from all 60 children. LDH
(730.08 6 452.63 m/L), Neu% (73.23 6 18.26), CRP
(66.36 6 33.96 mg/L) were increased during the acute
phase, while CD4 count (203.00 cells/mL; range: 2.00–
1560.00 cells/mL) and CD4/CD8 ratio (0.53 6 0.24)
was decreased.
Co-infections were detected in 26/60 (43.3%) chil-
dren. The results were C. albicans (n 5 11), aspergillus
(n 5 5), tuberculosis (n 5 4), Pseudomonas aeruginosa
(n 5 3), Acinetobacter baumannii (n 5 2) and Esche-
richia coli (n 5 1).
Part 2. Evaluation of CD4/CD8 as a biomarker
for PCP in pediatric patients
We collected CD4%, CD8% and CD4/CD8 ratio from
60 PCP-infected children.
We found that CD4% and CD4/CD8 ratio was sig-
nificantly decreased (17.90 6 6.73 vs 36.39 6 10.45, Figure 1. ROC curve analysis for CD4/CD8 as a marker for
0.53 6 0.24 vs 1.65 6 0.67, P < 0.05, P < 0.001, respec- identify PCP with other pneumonia.
study describes the clinical characteristics and outcome X-linked hyperimmunoglobulin M syndrome, and four
of non-HIV children with PCP at Beijing Children’s subjects with severe combined immunodeficiency dis-
Hospital from January 2005 to December 2014. ease (SCID). Of the 10 children, 7 died, which indicated
Hematologic malignancies and solid tumors are the that the occurrence of PCP in primary immunodefi-
most common underlying diseases in most of the study ciency children may have a poor prognosis. As such,
for non-HIV PCP-infected adults (3, 16, 17). However, appropriate prophylaxis should be considered in chil-
in our study, almost 2/5 of PCP children had inflam- dren with primary immunodeficiency disease.
matory diseases. This is consistent with the report from Glucocorticoid treatment is a well-known risk factor
Matsumura et al. (10). This may be because PCP pro- for PCP in non-HIV patients, and accounts for
phylaxis is common in hematological malignancies, but 55–97% of published cases (8, 9, 13, 18), and 66.7% in
not in inflammatory diseases, and spectrum of disease our study. The mechanism could be a decrease of
is different in adults compared to children. blood CD41 lymphocyte count due to glucocorticoid
Although primary immunodeficiency disease is a therapy (8). It is noteworthy that the duration of corti-
rare cause of PCP in adults, it occurred in 16.7% non- costeroid use prior to diagnosis of PCP was nearly
HIV children in our study. There were six subjects with 8 months in our study, while it was mostly 4–5 months
in adults (5, 13, 16, 19). During that time, children
often received follow-up large dose of glucocorticoid
Table 4. Multivariate analysis of independent factors
or some additional immunosuppressor because of
associated with mortality
underlying disease relapse. Therefore, routine PCP
Variables Odds ratio OR (95% CI) P value prophylaxis should be strongly considered during this
delay treatment 1.08 0.93–1.47 0.18 critical period.
LDH, U/L 2.57 1.23–6.71 0.04 Time from initiation of glucocorticoid medication
mechanical ventilation 0.96 0.77–1.45 0.04 to PCP diagnosis ranged from 14 to 2100 days. A
Coinfection 2.347 0.19–28.44 0.05 10-year-old boy with chronic idiopathic thrombocyto-
Pneumothorax 1.03 1.15–20.52 0.23
penic purpura (ITP) who received intermittent
glucocorticoid for seven years developed PCP after in time from onset to initial treatment between non-
another large dose of glucocorticoid administration. A survivors and survivors.
6-year-old boy with SLE developed PCP in 14 days Increased LDH level is likely correlated with wide-
after receiving methylprednisolone pulse therapy. As it spread tissue damage in PCP (11). Wide tissue dam-
reported in recent studies (5, 10), there is a wide range age reflects the severity of the illness. As previously
in the dosage and duration of treatment with glucocor- reported, high LDH levels should be regarded as a
ticoid in non-HIV before PCP development range. predictor of outcome in non-HIV children with PCP
Therefore, PCP prevention strategies in non-HIV chil- (5, 11).
dren will require considerations that account for indi- Co-infection may reflect that children with PCP
vidual differences. infection often have a low immune function caused by
The identification of biomarkers that can screen for both underlying disease and immunosuppressive med-
PCP risk in non-HIV children will be useful for early ications, which increases their risk for developing a
detection of this disease. Herein, CD4 counts were rou- variety of infections. We observed a high proportion of
tinely measured in PCP-infected children, with an subjects had co-infections (43.3%), with the most
average value of 203.00 cells/mL (range: 2.00–1560.00 common being C. albicans, tuberculosis and aspergillus.
cells/mL). Pyrgos et al. suggested that CD4 count below Non-HIV patients with PCP often develop fungal
250 cells/mL was a risk factor for PCP in children (4). infections (10, 22). We found 13 non-HIV children
However, CD4 counts >250 cells/mL were not rare in with fungal infections (C. albicans and aspergillus). C.
our PCP cohort. Furthermore, we did not find any dif- albicans infection rate reported here was higher than in
ference in CD4 levels in non-survivors and survivors in other studies in adults (8–10). Co-infection should be
our study (P > 0.05). Our study, therefore, suggests considered when there is a poor treatment response to
that CD4 counts may not be a suitable diagnostic and PCP therapy in non-HIV children.
prognostic marker in PCP infected non-HIV children. The need for mechanical ventilation was a poor
In our cohort, we collected CD4%, CD8% and prognostic factor in our study. Compared to HIV
CD4/CD8 ratio from 60 PCP-infected children. Their
patients with PCP, non-HIV patients are more prone
CD4/CD8 ratio was much lower than in non-PCP chil-
to ventilator-associated lung injury and the injury is
dren. We also assessed the possible utility of CD4/CD8
also more severe (6, 9, 17). In order to correct
ratio for identification of non-HIV children at a high
hypoxia, ventilator parameters are generally high.
risk of developing PCP. CD4/CD8 ratio was available
This often leads to ventilator-associated lung injury
in 60 non-HIV children, with a median ratio of
(23). In our study, we used high frequency oscilla-
0.53 6 0.24. Our results showed that using a cut-off
value of 0.715, the CD4/CD8 enabled PCP to be distin- tory ventilation (HFOV) in seven subjects, and
guished from other pneumonia with an AUC of 0.902 unfortunately, all of them died. Thus, we believe that
and an optimized sensitivity of 89.2% and specificity mechanical ventilation is a problem in PCP-infected
of 80.4%. We also found CD4/CD8 ratio decreased in children. Additional studies are required to improve
all infants younger than 1 year of age. Thus, the value protective lung ventilation strategies for this popula-
of the CD4/CD8 ratio is a suitable marker for predic- tion. Pneumothorax is an unsolved problem in PCP
tion and diagnosis of PCP in non-HIV children. at present. For adults, incidence of pneumothorax
To the best of our knowledge, mortality rate of PCP ranges from 4% to 36% (5, 6, 17, 24). In our study,
in non-HIV children has not been reported. In our pneumothorax rate was 26.7%. Nearly all of the
study, the mortality rate was 41.7%, which was similar non-HIV children who developed pneumothorax
to previous studies in non-HIV adults (17, 20–22). died. The combination of PCP-related lung injury
This indicates that PCP is also a life-threatening infec- and ventilator-associated lung injury may explain
tion in non-HIV children and there is a need for more the high occurrence rate in this cohort.
effective treatment to decrease its mortality. Our analy- This study has a number of limitations. First, retro-
sis determined that mortality was independently asso- spective studies, such as this, rely on existing patient
ciated with need for mechanical ventilation, LDH and records that were collected for medical reasons other
co-infection. than research, and when specific information is absent,
Time from onset to initial treatment has previously it is difficult to resolve. Fortunately, most variables in
been identified as a poor prognostic factor of PCP, and our study were the required items in the medical
there is a need for early intervention in order to records. Second, the follow-up beyond 30 days was not
improve survival (20, 21). However, when using multi- recorded, which does not allow an accurate prediction
variate analysis, our study did not detect any difference of long-term mortality.