How Useful can be CT Scan in the Differential Diagnosis of Pneumonia in Immunocompromised Children? The Experience at a Tertiary Care Pediatric Center

How Useful can be CT Scan in the Differential Diagnosis of Pneumonia in Immunocompromised Children? The Experience at a Tertiary Care Pediatric Center


Introduction
Etiological diagnosis of pulmonary infections in pediatric patients with hematologic malignancies or undergoing hematopoietic stem cell transplantation (HM-HSCT) can be challenging.According to guidelines, radiological findings, mostly identified at Computed Tomography (CT) scan, are required for diagnosis of invasive fungal disease with varying degrees of certainty [1,2], but they could be useful to suggest also a bacterial or viral etiology [3].CT scans have been demonstrated to be useful for early detection and follow-up (repeating imaging) of pulmonary invasive fungal disease (IFD), with important correlations with patients' outcomes [4,5].However, even if CT scans might have a great performance in the radiological identification of pneumonia in immunocompromised patients [6] and some findings could lead to a specific etiological diagnosis (e.g., halo sign and air crescent sign, suggesting IFD due to molds [7]), in most cases pulmonary lesions detected by CT scan are not specific in children [8][9][10].
Recently, CT pulmonary angiogram has been demonstrated useful in adults in identifying vessel occlusion sign (VOS), a radiological finding potentially helpful in the diagnosis of angioinvasive pulmonary aspergillosis [11][12][13][14].Other studies, instead, identified segmental consolidation as more specific of bacterial pneumonia [15].
In pediatrics, despite the utility of CT scan in immunocompromised patients, there could be concerns because of the radiation exposure [16,17].To overcome this issue, ultra-low doses CT scan was introduced in the last years, maintaining a good diagnostic performance [18].Furthermore, as for the possible risk of renal failure, a recent study demonstrated a low rate of acute kidney injury among pediatric population when contrast medium was used [19].
In this retrospective single-center study, we evaluated the role of CT scan in helping in etiological differential diagnosis of pneumonia in HM-HSCT children.

Patients and Methods
IRCCS Istituto Giannina Gaslini (IGG), Genoa (Italy), is a tertiary care children's hospital serving a region of 1.5 million inhabitants in the North-West of Italy, with a catchment area extended to the whole nation and many European and foreign countries for highly specialized medical and surgical pediatric care.
We retrospectively selected HM-HSCT patients admitted to IGG from February 1 st , 2009 to December 31 st , 2021with a chest CT scan performed and a defined infectious etiological diagnosis of pneumonia: fungal (according to EORTC-MSG criteria [2]), bacterial (in presence of lung infiltrate in a patient with bacteremia due to non-common skin contaminants or isolation of significant pathogens from bronchoalveolar lavage) or viral.We excluded patients with a diagnosis of pneumocystis pneumonia.
To derive the infectious etiological diagnosis (i.e., "definite diagnosis"), two infectious diseases specialist's expert in infections in immunocompromised hosts (EC and CR) reviewed the clinical records; based on clinical, microbiological and radiological findings, they classified the pulmonary infections as bacterial, fungal or viral."Definite diagnosis" was considered the gold standard diagnosis.
The CT scans of the selected patients were retrospectively evaluated by one senior and one young radiologist, blinded to the "definite diagnosis", who performed two separate readings of all CT scans and proposed a diagnosis based on only radiological findings (i.e., "radiological diagnosis").To reduce the risk of outcome bias, the two readings made by the same radiologist were performed about 1 month apart, and the senior radiologist designated to CT reading was selected if a period greater than 5 years had passed from the initial CT report to the beginning of this study.
The following clinical data obtained from clinical records were recorded: age, sex, hematological disease, and type of bone marrow transplantation.
The primary aim of the study was to identify radiological findings that alone allow for differentiating fungal, bacterial and viral pneumonia in HM-HSCT patients.

Secondary aims were:
-to evaluate the "radiological" and "definite diagnosis" concordance -to evaluate the "intra-radiologist concordance", that is the concordance in radiological findings interpretation in 2 serial readings performed by the senior and the junior radiologists, comparing senior first (S1) versus senior second (S2) reading and junior first (J1) versus junior second (J2) reading; -to evaluate "inter-radiologist concordance", that is the concordance in radiological finding interpretations between senior and junior radiologists, comparing S1 vs. J1 and S2 vs. J2.

Image Analysis
CT scans performed in HM-HSCT patients with an infectious etiological diagnosis of pneumonia were collected and anonymized.A Siemens Somaton 64 slices CT scanner was used; the same examination protocol was applied for all readings: volumetric thoracic scan, 3 mm MPR reconstruction with lung and mediastinal windowing was applied.
Based on these results, a CT scan reading table ( [39]."Halo sign" (Figure 2a) was defined as ground-glass opacity surrounding a pulmonary nodule or mass; "air-crescent sign" (Figure 1a) as the presence of air in a crescent shape in a nodule or mass, and "reversed halo sign" as a focal rounded area of ground-glass opacity surrounded by a crescent or complete ring of consolidation.Radiologists, one senior (BD, Pediatric Radiologist Physicians) and one junior (VS, Radiologist Resident), blinded for patients' clinical characteristics and the "definite diagnosis", performed two readings of every CT scan using the reading table.For each item in Table 1, the assessment options were "yes" when the radiologists identified the radiological sign, otherwise, they had to choose between "no" or "not assessable".

Consolidation
For each reading, radiologists had to hypothesize a diagnosis (bacterial, fungal, viral, or not identifiable infection), called "radiological diagnosis", based on imaging findings only.To derive the "radiological diagnosis" they relied on Table 2 [18].At the end of the readings and diagnostic radiological hypothesis process, for each CT scan we had 4 "radiological diagnoses" (derived from first [S1] and second [S2] senior readings and from first [J1] and second [J2] junior readings) and one "definite diagnosis".

Fungal pneumonia
Nodules or patch areas of consolidation with a halo of surrounding ground-glass opacity cavitation or lung ball (late phase); or focal rounded area of ground-glass opacity surrounded by a crescent or complete ring of consolidation (reversed halo sign).

Viral pneumonia
Mosaic attenuation pattern (patchy areas of inhomogeneous lung attenuation caused by hyperventilation of alveoli distal to bronchiolar obstruction), patchy and poorly-defined areas of consolidation or bilateral patchy areas of GGO along broncho vascular bundles or along subpleural lungs in both lungs with random distribution, or bilateral lesions of centrilobular small nodule with short-branching pattern showing tree-in-bud signs CT: computed tomography; GGO: ground-glass opacity Table 2: Radiological appearance of different infectious diseases in immunocompromised hosts [18]."Radiological diagnosis" was then compared with "definite diagnosis" and, based on the final diagnosis concordance, the "radiological diagnosis" with higher concordance with "definite diagnosis" was chosen and a sub-analysis of all radiological items included in the table was performed, trying to identify the single specific radiological finding that could allow to distinguish between different infectious pneumonia.

Statistical analysis
Descriptive statistics were generated for the whole cohort and data were expressed as mean and standard deviation for continuous variables.Median value and range were calculated and reported, as were absolute or relative frequencies for categorical variables.
Non-parametric analysis (Mann-Whitney U-test) for continuous variables and Chi square or Fisher's exact test for categorical variables were used to measure differences between the groups.
A P value less than 0.05 was considered statistically significant, and all values were based on two-tailed tests.Statistical analysis was performed using SPSS for Windows (SPSS Inc., Chicago, Illinois USA).
According to Italian legislation, the study did not need ethical approval since it was a purely observational retrospective study and therefore it was not possible to request informed consent for participation.In any case, consent to completely anonymous use of clinical data for research/epidemiological purposes is requested by the clinical routine at the time of admission/diagnostic procedure.

"Radiological diagnosis" and "definite diagnosis" concordance
The frequency of different etiological diagnoses based on "radiological diagnosis" and "definite diagnosis" is reported in Table 4.In 2 cases the initial CT scan report was expressed by the senior radiologist, 5 and 7 years before this study was started.Senior radiologist classified the CT scans as non-specific/negative in 4 cases in the first and in 5 cases in the second reading; no CT scan was suggestive of viral pneumonia according to senior radiologist's opinion.
Both small and big 1 (  Since the primary outcome was to identify radiological findings useful in defining the infectious etiology of pneumonia in HM-HSCT population, we chose the "radiological diagnosis" with the higher agreement with "definite diagnosis" (S1) for this evaluation, subsequently analyzing the frequency of different radiological findings in bacterial and fungal pneumonia, as defined by the radiologist in S1 "radiological diagnosis" (Table 7).
Almost none of the radiological findings was specific for fungal or bacterial pneumonia (Table 6).Particularly, consolidation was present in the same proportion in patients with radiological diagnosis of bacterial and fungal pneumonia, but in bacterial pneumonia it was more frequently multifocal and lobar, while in fungal pneumonia focal, even if statistical significancy was not reached.Halo sign was found more frequently in fungal pneumonia, while nodules were represented in 50% of both bacterial and fungal pneumonia, but they resulted more frequently single and cavitated in fungal pneumonia, multiple in bacterial pneumonia.Ground glass opacities were multifocal and central in most cases of bacterial pneumonia, peripheral in fungal pneumonia.Presence of vessel occlusion sign and aneurysm/pseudoaneurysm were not useful in differential diagnosis between bacterial and fungal pneumonia.

Discussion
In the present study, we evaluated the contribution of CT scan in the differential diagnosis of pneumonia in HM-HSCT children with radiologists blinded to clinical and microbiological data.In this condition, the 1 st "radiological diagnosis" made by a senior radiologist had the best concordance with the "definite diagnosis", but none of the radiological findings were significantly associated with diagnosis of bacterial or fungal pneumonia (Table 7), confirming the lower specificity of imaging in pediatrics, compared with adults [41].For example, halo sign, that in adults is considered highly specific for invasive pulmonary aspergillosis, in our cohort was identified not only in 41.7% of CT scans of patients with a diagnosis of IFD, but also in 16.7% of those with a diagnosis of bacterial pneumonia.Moreover, cavitation was observed in the presence of both infections, but in this case the presence or absence of neutropenia, a "clinical" datum, represented the driver for the definite diagnosis, with cavitary lesions in neutropenic patients not related with invasive mycosis [42].Finally, vessel occlusion that has been suggested as a sign of mold pneumonia in adults [12][13][14]21] was observed only in 12.5% of children with fungal pneumonia and 25% of those with bacterial infection.
Another interesting observation was that radiologists are consistent with their own readings, with intra-radiologist concordance higher for junior radiologist (J1 vs. J2 "radiological diagnosis") than between senior (S1 vs. S2).A possible explanation for the higher intra-observer concordance in J1 vs. J2 is that junior radiologists tend to be more adherent to scientific and academic descriptions of radiological signs, which are also more reproducible, while senior radiologists rely not only on academic definitions, but also on personal experience.These might also explain the substantial agreement between senior "radiological diagnosis" and "definite diagnosis", which was lower comparing junior "radiological diagnosis" to "definite diagnosis", and consequently the low inter-observer agreement.Looking at Table 3, we can also hypothesize that senior radiologists were not only more confident in identifying fungal and bacterial radiological findings (compared to "definite diagnosis") but were also more confident in saying that CT scans were non-specific in some cases.On the contrary, junior radiologist was probably less confident in defining the imaging as non-specific, classifying 5 and 4 CT scans in J1 and J2 readings, respectively, as viral.
Intriguingly, the concordance between 1 st senior radiologist diagnosis and "definite diagnosis" was substantial (Kappa 0.69, 95%CI 0.45;0.93.Agreement 83.3%,), while the 2 nd had a lower agreement, suggesting that "a second look" could have entered doubts even in senior radiologist if not sustained by clinical, laboratory and microbiological data.
The major limitation of our study is the small sample size (23 patients over a period of 12 years), which could be explained by the careful application of antifungal prophylaxis protocol.Due to this issue, it is difficult to identify radiological findings able to diversify with statistical significance between bacterial or fungal pneumonia.However, a senior radiologist, who has expertise in pediatric and immunocompromised patients, was able to draw up a "radiological diagnosis" concord with "definite diagnosis" in 83.3% of cases, highlighting a good radiological diagnostic performance.Moreover, the blinded readings of CT scans made by two groups of radiologists increase the importance of our findings and at the same time underlines the need for cooperation between radiologists and clinicians' experts in the management of infections in HM-HSCT pediatric patients.Indeed, the low specificity of radiological findings in this specific population makes it necessary to combine imaging with clinical, laboratory and microbiological data [24,37] to come to a correct diagnosis.
In conclusion, our study confirms that CT scan findings in pneumonia in HM-HSCT children are frequently not specific.What is in our opinion the most important observation is that CT findings must be interpreted in the light of clinical, laboratory and microbiological data, meaning the need of strict cooperation between radiologists and clinicians taking care of infections in immunocompromised children.This cooperation among different specialists must be extended to teaching younger medical doctors to provide the best care, especially in complex patient populations such as HM-HSCT children.

Figure 1 :
Figure 1: Fungal infection.Volumetric CT post contrast medium.MPR thin (3 mm) coronal reconstruction in a 10-years-old boy with acute T-cell leukemia, neutropenia, and persistent respiratory symptoms.

Figure 1A :
Figure 1A: Lung windowing: superior left lobe parenchymal consolidation (arrow) with air in a crescent shape within the nodule (air crescent sign).Small (<10mm) nodule with cavitation within subpleural region of inferior left lobe (short arrow).

Figure 1B :
Figure 1B: Mediastinal windowing: central hypodensity area (arrow) as result of a central area of necrosis (hypodense sign).Pleural effusion is evident in left lateral costophrenic recess (short arrow).

Figure 2 :
Figure 2: Aspergillosis.Volumetric CT post contrast medium.MPR thin (3 mm) axial reconstruction in a 14-years-old girl with bone marrow aplasia, neutropenic, who underwent HSCT five months before pulmonary infection.

Figure 2A
Figure 2A Lung windowing: focal subsegmental superior parahylar right consolidation (arrow).A halo of surrounding ground-glass opacity, halo sign, is evident around the lesion.

Figure 2B
Figure 2B Mediastinal windowing: central hypodensity area (arrow) within the lesion as a result of a central area of necrosis (hypodense sign).

Figure 3A
Figure 3ALung windowing: multiple diffuse bilateral areas of segmental and subsegmental consolidations involving central and peripheral portions of lungs.Also, diffuse bilateral areas of ground-glass opacities.

Figure 3B
Figure 3BMediastinal windowing: multiple diffuse bilateral areas of segmental and subsegmental consolidations involving central and peripheral portions of lungs.Also, diffuse bilateral areas of ground-glass opacities.

Figure 4 :
Figure 4: Bacterial pneumonia.9-years -old boy with acute B-cell leukemia, fever, and desaturation, who underwent HSCT seven months before pulmonary infection.

Figure 4A
Figure 4A Volumetric CT MPR thin (3 mm) coronal reconstruction.Lung windowing: bilateral areas of multifocal segmental/ subsegmental consolidations at the level of the carina.

Figure 4B
Figure 4B Volumetric CT MPR thin (3 mm) axial reconstruction.Mediastinal windowing: multifocal and multilobar distribution of the areas of consolidations with a prevalent central distribution at subcarinal level.

Figure 5 :
Figure 5: Bacterial pneumonia.Volumetric CT MPR thin (3 mm) coronal reconstruction in a 19-months-old boy with acute myeloid leukemia, persistent fever, and respiratory symptoms.

Figure 5A
Figure 5A Lung windowing: focal lobar consolidation (star) with a hypodense area within the lesion as a result of a central area of necrosis (hypodense sign).

Figure 5B
Figure 5BMediastinal windowing: focal hemorrhagic area within the lesion (star) associated with a cavitation area.Posterior contralateral basal dystelectasic consolidation is also evident.

Table 1 :
Radiological findings looked for by radiologists in CT scans.
Table 3 reports on clinical characteristics of detected cases.

Table 4 :
Frequency of different etiological diagnosis according to the four "radiological diagnosis" and "definite diagnosis".

Table 5 :
Radiological diagnosis and concordance among different readings and definite diagnosis.

Table 7 :
Frequency of different radiological findings at 1st reading made by senior radiologist (S1) in "definite diagnosis" of bacterial and fungal pneumonia in children with hematologic malignancies or undergoing hematopoietic stem cell transplantation.Russo C, Rizzo F, Sambuceti V, Anfigeno L, Calevo MG, et al., (2023) How Useful can be CT Scan in the Differential Diagnosis of Pneumonia in Immunocompromised Children?The Experience at a Tertiary Care Pediatric Center.Infect Dis Diag Treat 7: 243.DOI: 10.29011/2577-1515.100243 Citation: