Pulmonary lesions in early response assessment in pediatric Hodgkin lymphoma: prevalence and possible implications for initial staging

Hodgkin lymphoma (HL) is a rare cancer in children under 15 years of age, but it is the most common cancer among adolescents 15–19 years of age [1, 2]. Treatment of HL has become a great success in modern oncology, with survival rates exceeding 90% for patients irrespective of the stage of disease [3‐10]. Survivors of classical Hodgkin lymphoma (cHL) are at high risk of secondary malignancies and cardiovascular disease [1, 11‐17]. Increased risks persist for decades after treatment and are clearly correlated with the extent of treatment exposure [18].

Disease staging is most frequently based on the Ann Arbor classification [19] with Cotswold modifications [20]. Nowadays, modern staging systems combine anatomic cross-sectional imaging, including computed tomography (CT) and/or magnetic resonance imaging (MRI), with functional positron emission tomography (PET) [21]. The Lugano criteria published in 2014 are the most recent updated consensus of the International Conference on Malignant Lymphomas Imaging Working Group [22, 23].

Disseminated organ involvement of the lung constitutes stage IV disease. Thus, patients are preassigned to a high-risk stage regardless of other prognostic factors. They receive multi-agent chemotherapy with response-adapted radiotherapy restricted to patients with residual disease on interim staging fluorodeoxyglucose (FDG)-PET scans. Compared with their age-matched controls, after lung radiotherapy with 15 Gy to ≤ 25 Gy, patients have an increased risk of lung fibrosis and recurrent pneumonia [18, 24]. The challenge is to maintain high cure rates while reducing adverse therapy-related side effects. Therefore, the role of imaging is critical to differentiate real pulmonary metastases of HL from other causes of lung lesions common within a pediatric population.

The differential diagnosis of pulmonary abnormalities in patients with HL includes a variety of diseases such as infection (e.g., histoplasmosis or tuberculosis), bronchiolitis obliterans, organizing pneumonia, granulomatosis with polyangiitis, drug toxicity, and effects of radiotherapy [25‐29]. Given that pediatric Hodgkin lymphoma (pHL) is an immunocompromising disease [30] that facilitates benign lung changes, any given patient may have a mixture of pulmonary entities. Even in the case of histologically proven malignancy, it is impossible for radiologists to reliably distinguish benign from malignant lesions on a single scan [31, 32]. However, this may be possible with hindsight if the behavior of the individual nodule and the patient’s overall outcome are known.

A lung biopsy for definitive staging of lung nodules is challenging due to the low sensitivity and specificity of a lung biopsy and the morbidity of this procedure. Therefore, given the limited ability of imaging to determine whether a single lung lesion is a metastasis or background noise, a pragmatic approach must be adopted. If lung lesions are found without clinical signs of infection and respond to chemotherapy like extrapulmonary, histologically proven manifestations of HL, pulmonary involvement must be assumed.

Since therapeutic decisions are based on initial staging, a retrospective approach looking at how lung lesions respond to therapy to determine their etiology is not feasible. At present, two of the largest cooperative study groups in Europe and North America already take this dilemma into account in their staging definitions of lung involvement: the European Network for Pediatric HL relies entirely on CT-morphologic criteria including size and number of pulmonary lesions, whereas the Children’s Oncology Group from North America also includes metabolic factors (Table 1). The size criteria for all study protocols are not based on published original research data but on consensus.

Current international collaborative efforts to harmonize the staging evaluation and response criteria in clinical trials for children, adolescents, and young adults with HL are ongoing [33].

As a next step in distinguishing true pulmonary metastases from background noise on initial staging, parameters such as size, number, and morphologic criteria must be considered. One way to solve the problem is to compare the initial imaging with the interim staging during therapy in patients with no recurrence or progression of disease. These new pulmonary lesions might be the result of other causes, such as opportunistic infection due to chemotherapy-induced neutropenia or immunosuppression from HL itself.

Thus, the description of these lesions represents a good approximation of the background noise that complicates the delineation of true pulmonary metastases to initial staging. This distinction is urgently needed to prevent patients from erroneous upstaging to stage IV and for tailoring therapeutic strategies.

This study attempts to:

This retrospective analysis included patients enrolled in the EuroNet-PHL-C1 trial (EudraCT: 2006–000995-33; Clinicaltrial.gov: NCT00433459). This trial recruited 2,102 pHL patients younger than 18 years with cHL between January 31, 2007, and January 30, 2013. The imaging protocol comprises a CT of the chest for initial staging and for interim staging defined as early response assessment (ERA) on days 29 to 31 of the second cycle of chemotherapy consisting of vincristine, etoposide, prednisolone, doxorubicin (OEPA; 1–5 mg/m² vincristine intravenously capped at 2 mg, on days 1, 8, and 15; 125 mg/m² etoposide intravenously on days 1–5; 60 mg/m² prednisone taken orally on days 1–15; and 40 mg/m² doxorubicin intravenously on days 1 and 15) [9, 10].

Written informed consent was provided by all patients or their guardians. Local and central ethics boards in each country approved the study, which was conducted in accordance with Good Clinical Practice and the Declaration of Helsinki.

Out of 2,102 study participants in the EuroNet-PHL-C1 trial, 1,752 (83.3%) were available for central review. Incomplete or missing images to the central review board and inadequate imaging quality were imaging-related exclusion criteria (Fig. 1). Inadequate imaging quality was defined as significant respiratory or motion artifacts (apparent gaps in lung coverage), slice thickness equal or larger than 10 mm, or low spatial resolution on ultra-low-dose CTs or low-dose CTs acquired only for attenuation correction of PET. This retrospective analysis includes CT examinations of 1,300 patients. Each patient’s initial staging and ERA images were reviewed for new lung lesions that developed during two cycles of OEPA. While PET/CTs are utilized for staging and response assessment [22, 23], this analysis was done without regard to metabolic activity of the lung lesions due to the frequent unreliability and of lack of FDG uptake in small lung nodules (< 1 cm) [34].

To evaluate the presence of pulmonary lesions, three physicians (D.S., radiologist, 15 years of experience; J.S., resident in radiology, 2 years of experience; C.L., specialist in internal medicine and pneumology; 7 years of experience) reviewed CT scans at standard lung window settings (center, -600 HU; width, 1,500 HU). For each examination, the reconstruction parameters of slice thickness, increment, and field of view (FOV) were recorded.

A comparison was made between the appearance of abnormalities on ERA and those on initial staging. New pulmonary lesions on ERA imaging were classified as one of the following patterns based upon the currently recommended terms of the Fleischner Society [35] and previously published data on lung involvement in pHL [25, 36] (Fig. 2): (1) nodules, (2) masses, (3) ground-glass opacities, (4) consolidations, (5) parenchymal bands, and (6) perilymphatic distribution.

After defining the pattern of a lung lesion, the following parameters were recorded:

Documentation of newly developed lung lesions was restricted to ten per patient to limit the impact of a single patient on the pattern-based analysis. If different types of lesions were present, we included them all in a ratio that reflected their frequency in the patient. Data were processed with respect to progression or relapse over a 5-year period. Changes in mediastinal or hilar lymphadenopathy over time were assessed in patients without relapse or progression of disease but with new lung lesions on ERA. The statistical analysis was conducted using R (The R Foundation for Statistical Computing, Version 4.0.2., Vienna, Austria).

Disseminated pulmonary involvement of HL is indicative of stage IV disease which requires a more intensive chemotherapy regimen and/or radiotherapy. Correct imaging and appropriate interpretation of abnormalities are even more important for children and adolescents to reduce the risk of secondary malignancies, infertility, and cardiovascular disease and to preserve quality of life [37]. Considering that HL is an immunocompromising disease that facilitates benign lung changes, it is very often unclear if detectable lung lesions are part of the malignancy or non-malignant confounders representing a background noise. The differentiation of these two entities is a prerequisite for an adequate therapy decision.

Biopsy of lung lesions is not feasible in a research context for ethical reasons and because of its low diagnostic yield [38, 39]. Consequently, this paper analyzed new lesions on ERA, because it can be assumed that new lesions developing under chemotherapy that are not associated with relapse or progression can be considered benign in nature and thus most likely represent an approximation of this background noise. There are no published data available on newly developed pulmonary lesions on follow-up images for pHL patients.

There were 119 out of 1,300 (9.2%) patients that presented with new pulmonary lesions on ERA. The phenomenon occurs regardless of whether or not lung involvement was already present at the time of initial staging or even regardless of whether a patient develops relapse. In the latter group, new lung lesions at the time of ERA regressed by the time of relapse staging. Therefore, we assume that new lesions on ERA require no further follow-up imaging because lung progression is extremely rare on ERA.

The majority of newly developed lung lesions were nodules. Consolidations and ground-glass opacities were less common. Many patients presented with a single new nodule, but a minority also developed multiple lesions. Almost all nodules were smaller than 10 mm. No masses were detected. The right lung and the basal lung segments are more likely to be affected which might be explained by lung volume [40] and perfusion [41].

Non-malignant causes of new lung lesions on ERA are diverse. In addition to direct or indirect therapeutic effects, infectious causes may also be considered [42]. No direct pulmonary toxicity has yet been described for the OEPA combination chemotherapy and in accordance with the treatment regimen, there was no radiotherapy between the two points in time of the examinations [9, 10]. It is therefore assumed that lung lesions that developed between initial staging and ERA within approximately 2 months were not caused by direct chemotherapy effects. It is therefore more likely that they are due to opportunistic infections favored by the neutropenia or leukopenia caused by chemotherapy as an indirect effect of OEPA [10, 43] or to the immunosuppression caused by the underlying disease itself [30]. The second mechanism mentioned is a circumstance that is more pronounced in initial staging, as disease activity typically decreases during therapy.

Rosenfield et al. found that even with a known underlying malignancy, approximately 1/3 of biopsied pulmonary nodules are found to be benign. Etiologies included granulomatous disease, infection, inflammatory myofibroblastic lesion, drug reaction, scarring, and intrapulmonary lymph nodes [31]. The Fleischner criteria [44] are not suitable as they relate to incidental pulmonary nodules. Furthermore, they do not comply with patients younger than 35 years, immunocompromised patients, or patients with known cancer.

In our study, no progressive hilar or mediastinal lymphadenopathy was present in any of the patients with new pulmonary nodules; thus, hilar lymphadenopathy does not seem to be a factor to identify infectious causes of new lung nodules.

On ERA, 72 of 1,300 patients (5.5%) had new nodules in the lung mostly<10 mm in size. Maturen et al. found pulmonary involvement in 12% in HL on initial staging with nodules present in 90% of the lung involvement cases [36]. It is morphologically impossible to distinguish lung lesions from lymph nodes in some cases since hilar lymph nodes are present on a lobar, segmental, and subsegmental level [45]. Hilar lymph nodes have been described in up to 60% of the patients in a cohort of 120 children aged 1–17 years who underwent emergency CT after high-energy trauma. They measured up to 9 mm [46].

Pulmonary nodules up to 10 mm in size of benign etiology are therefore a common occurrence in pediatric and adolescent patients with HL.

Most patients had only one newly developed nodule on ERA, but the distribution of patients with two to ten new nodules was relatively homogeneous, suggesting that the number of pulmonary nodules does not appear to be a reliable staging parameter for defining lung involvement. Location and border configuration of the nodules do not seem to be relevant parameters for distinguishing benign from HL related nodules since the benign nodules in our collective are evenly distributed in both lungs and all border configurations occur. The predominance of poorly defined borders is discussed in the limitation section.

In contrast to nodules, newly developed consolidations and ground-glass-opacities present a greater challenge in interpretation because information about antibiotic therapy was not recorded on all staging forms in these patients. In the study by Diederich et al. [25], consolidations on CT before initiation of therapy were detectable in 27% of patients with lung changes in HL, a similar frequency as in our cohort on ERA. Consequently, it is possible for consolidations to be both the result of pulmonary metastatic disease and non-malignant confounders, mainly opportunistic infections. Unfortunately, we were not able to identify a distinguishing morphological feature in our cohort with benign consolidation and ground-glass opacities. Newly developed consolidations in ERA with most likely reactive lymphadenopathy were present in 4 out of 29 patients but hilar or mediastinal lymphadenopathy does not seem to be a reliable parameter for initial staging, as it is common in pHL. As a result, the inclusion of PET, laboratory, and clinical parameters is critical for differentiation.

Although new cavitations on ERA (Fig. 7) were rare, their genesis can be traced from the patient record in both cases: one was a 4-year-old boy who was treated with caspofungin and amphotericin B. After 14 days of antifungal therapy, the lesion fully resolved. A 13-year-old girl was also treated for fungal infection, but no further images were available for central review. In general, fungal infections are not uncommon in an immunocompromised population and should be considered as potential differential diagnosis, but are rare in pHL given the mild degree of immunosuppression compared to chemotherapy for other diseases.

In one patient with a newly developed consolidation, pulmonary artery embolism was present on ERA, so the consolidation should be diagnosed as infarct pneumonia. This cause of consolidations should be considered, especially in tumor patients, although many staging CTs are not feasible due to the contrast agent phase in our collective 39% of the ERA CTs were performed without contrast agent.

Parenchymal bands are usually due to distortion of the lung parenchyma [35]. Newly appearing parenchymal bands on ERA may be due to uneven re-expansion of the lung as the mediastinal tumor conglomerate shrinks. The small number of patients does not allow any further conclusions.

A perilymphatic distribution was not recorded, which could be explained by the decrease in size of the mediastinal mass and thus the removal of the cause of lymphatic congestion.

Although differential diagnostic uncertainty remains with regard to the detectable lesions, an isolated pulmonary progression after two cycles of chemotherapy that spontaneously resolves during further therapy is theoretically possible, but seems very unlikely in HL. CT remains the best option especially since invasive procedures do not provide sufficient sensitivity [31, 38] and are not a suitable alternative from an ethical perspective. As other publications already suggest [32, 47], it is most likely that not every lung lesion detectable on initial staging represents a HL manifestation, but that there is a coexistence of non-malignant and malignant lesions. Consequently, newly developed lesions represent the closest possible approximation to the non-malignant background noise on initial staging. Since the main effort of pediatric HL trials in the past decades was to reduce overtreatment, it may be necessary to rethink the definition of lung lesions on initial staging. Based on our findings in this study and the literature on benign lymph nodes in children [46], we recommend that lung nodules should only be considered disseminated lung involvement if they are larger than 10 mm in diameter. In contrast to the current staging criteria of EuroNet-PHL-C1 and C2, the number of lesions seems to be less important.

In contrast to the current staging protocol for lung nodules, this would allow a proportion of patients with several but small lung lesions to be allocated in lower treatment groups. It would also categorize patients with malignant lung lesions <10 mm into lower treatment groups, but this approach is already common practice established for the definition of involved lymph node regions [33]. Future studies need to closely examine the outcome of these patients. Therefore, nodules <10 mm should still be documented.

In this study, our limited data on cavitations suggests that all cavitations on initial staging should be screened for fungal and other possible infectious causes. The presence of hilar and or mediastinal lymph nodes is not a useful parameter on initial staging since most patients have mediastinal and or hilar involvement.

One major limitation of the approach used in this study to extrapolate phenomena observed at ERA compared to the initial staging is that the degree of immunosuppression at ERA would be more significant compared to that at initial diagnosis, given that chemotherapy was administered placing patients at higher risk for infection. This may result in more subcentimeter nodules due to infectious etiologies, chemotherapy toxicity, etc. In effect, this would overestimate the background prevalence of subcentimeter nodules in pHL patients who are not receiving chemotherapy (i.e. at initial diagnosis).

Other limitations of the study were mainly caused by the high number of participating study centers in this multinational trial, the retrospective study design, and absence of a strict protocol to which participating centers adhered. This resulted in non-standardized protocols being used and therefore in image data-related problems such as low spatial resolution due to an oversized FOV and an excessive slice thickness in a significant number of cases. Especially in small nodules, border configuration can be afflicted by the mean slice thickness of 3.3 mm in our study. The Fleischner Society recommends that for nodule characterization and measurement of small pulmonary nodules, CT scans should be reconstructed with contiguous thin slices (1.5 mm or less in thickness) [44]. Consequently, the number of nodules with poorly defined borders is overestimated in our study. A slice thickness of 1.5 mm or less and a FOV adapted to the patient’s thoracic size should therefore be implemented into the imaging protocols of future trials. Furthermore, it is recommended to use a uniform terminology based on the current Fleischner Society recommendations [35] in order to facilitate the comparison of studies.

A significant proportion of patients recruited in the EuroNet-PHL-C1 trial had to be excluded from the analysis due to unavailability or incompleteness of imaging which reduces the study population and therefore the reliability of the results. The image evaluation and pattern classification of the lesions were done by consensus, which could have influenced the results of the study and its reproducibility.

Another limitation is the missing control group of pHL patients with histologically proven malignant lung lesions. As discussed in the “Introduction” of this paper, lung biopsy is not feasible to distinguish benign from malignant lesions. It has to be assumed that both types of lung lesions coexist in pHL patients on initial staging. Therefore, this study only describes the characteristics of the background noise and is limited in describing discriminating characteristics of benign and malignant lesions.

Children and adolescents with HL often develop new pulmonary lesions between initial staging and ERA. If there is no progression in lymph node regions on ERA, these lung lesions should not be considered as progression and therefore do not require any further action. While newly developed nodules measured up to 36 mm, the majority of them (96.3%) were smaller than 10 mm. As the favoring conditions for the development of these lesions are also present at the time of initial staging, it must be assumed that benign and malignant lung lesions coexist before starting therapy. To reduce upstaging to stage IV and the associated overtreatment, this background noise needs to be distinguished from true pulmonary metastases. Raising the cut-off size of lung nodules to ≥ 10 mm to define involvement will reduce overtreatment, but this definition needs additional evaluation with outcome data. In contrast to the staging criteria of EuroNet-PHL-C1 and C2, our data suggest that the number of lesions present at initial staging may be less important.