Chest CT Diagnosis and Clinical Management of Drug-related Pneumonitis in Patients Receiving Molecular Targeting Agents and Immune Checkpoint Inhibitors: A Position Paper from the Fleischner Society

R advances in cancer biology have opened a new era of molecular targeting therapy, including tyrosine kinase inhibitors (TKIs) targeting epidermal growth factor receptor (EGFR) that have a significant treatment advantage over cytotoxic agents for the treatment of advanced non–small cell lung cancer (NSCLC) (1,2). More recently, an entirely new group of drugs has been shown to be beneficial to patients with advanced-stage cancer (3–6). This includes immune checkpoint inhibitors (ICIs) that prolong survival and improve quality of life in patients with advanced-stage cancers including NSCLCs (7). In a U.S. cross-sectional study (8), patients with cancer eligible for ICIs increased from 1.54% in 2011 to 43.63% in 2018; patients’ response to ICIs was 0.14% in 2011 and increased to 12.46% in 2018. Since the first report from Japan of severe acute lung injury in patients with NSCLC treated with gefitinib (9), pneumonitis associated with molecular targeting agents has attracted considerable attention. More recently, ICIs such as nivolumab and pembrolizumab have demonstrated to be associated with toxicities often termed immune-related adverse effects, including pneumonitis as one of the clinically significant and potentially life-threatening toxicities (10–12). The availability of serial chest CT Chest CT Diagnosis and Clinical Management of Drug-related Pneumonitis in Patients Receiving Molecular Targeting Agents and Immune Checkpoint Inhibitors: A Position Paper from the Fleischner Society


Clinical Features of DRP
The incidence of DRP varies widely among published studies. A recent population-based study from France (25) estimated an incidence of 1.2 per 100 000 per year and a prevalence of 2.6 per 100 000. DRP accounts for 2.5%-5% of prevalent cases of ILD (25)(26)(27). Cancer drugs (eg, bleomycin) are the most common cause of DRP, followed by drugs for autoimmune diseases (eg, methotrexate), amiodarone, and antibiotics (eg, nitrofurantoin), based on a recent systematic review (27). Additionally, awareness of the incidence and risk factors of pneumonitis related to specific anticancer agents is increasing in importance.
DRP can manifest as a variety of clinical syndromes of lung injury (27)(28)(29)(30)(31). The onset of illness may be acute, insidious, and sometimes delayed with a long latent period (eg, beyond 10 years in some cases of carmustine-induced pulmonary fibrosis) after completion of drug treatment (27,28). The clinical symptoms are generally nonspecific, including dyspnea, cough, malaise, and low-grade fever. Some patients may be asymptomatic even in the presence of diffuse pulmonary opacities. Lung auscultation often reveals crackles but may be normal. It is difficult to clinically distinguish DRP from lung disease of other causes such as infections, pulmonary hemorrhage, pulmonary edema, radiation-induced pneumonitis, and metastases. In addition, evaluation for cardiovascular etiologies including heart failure, Abbreviations DAD = diffuse alveolar damage, DRP = drug-related pneumonitis, EGFR = epidermal growth factor receptor, HP = hypersensitivity pneumonitis, ICI = immune checkpoint inhibitor, ILD = interstitial lung disease, NSCLC = non-small cell lung cancer, NSIP = nonspecific interstitial pneumonia, OP = organizing pneumonia, PD-1 = programmed cell death protein 1, PD-L1 = programmed death ligand 1, TKI = tyrosine kinase inhibitor Summary Increasing use of molecular targeting agents and immune checkpoint inhibitors has increased the frequency and broadened the spectrum of lung toxicity, particularly in patients with cancer. In this position paper from the Fleischner Society, the authors provide diagnostic criteria and management recommendations of drug-related pneumonitis for radiologists, clinicians, clinical trialists, and trial sponsors.
Essentials n The CT patterns in drug-related pneumonitis (DRP) reflect acute (diffuse alveolar damage) interstitial pneumonia and transient (simple pulmonary eosinophilia) lung abnormality, subacute interstitial disease (organizing pneumonia and hypersensitivity pneumonitis), and chronic interstitial disease (nonspecific interstitial pneumonia).
n The diagnostic criteria include newly identified pulmonary parenchymal opacities at imaging, temporal association of presentation with the initiation of a systemic therapeutic agent, and the exclusion of other likely causes.
n Management of DRP consists of drug discontinuation, immunosuppressive therapy, or both, along with supportive measures including supplemental oxygen and intensive care.
scans during cancer treatment allows detailed evaluations of both presymptomatic and symptomatic drug-related pneumonitis (DRP) and characterization of its radiologic manifestations in patients with different cancer types treated with different agents. The reports on pneumonitis among patients with NSCLC receiving EGFR-targeted TKIs, or EGFR-TKIs, included acute fatal events (13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23). Moreover, the recent introduction of novel molecular targeting agents such as third-generation EGFR-TKIs, newer anaplastic lymphoma kinase inhibitors, BRAF inhibitors, antibody-drug conjugates, and use of ICI combination regimens has brought new challenges to DRP (24). Despite many reports on this topic, there is limited published guidance on the evaluation and treatment of patients with DRP secondary to treatment with ICIs.
In this position paper from the Fleischner Society, we identified specific questions for diagnosis and management of DRP in patients receiving molecular targeting agents and ICIs. We conducted a systematic search to identify supporting evidence on these topics that had been published since the introduction of EGFR-TKIs in 2003. Based on the review of the literature combined with expert opinions, this position paper provides a description of radiologic patterns; a set of proposed diagnostic criteria; and management recommendations for DRP, which will be of value in improving the understanding of the pathophysiology, prognosis, and management of this diverse group of conditions. We believe that this position paper will be of interest to a diverse group of health care providers and clinical researchers. 3 pected outcomes for the individual patient. These issues need to be discussed with the patient in a shared decision-making process that incorporates the individual patient's values and preferences.
An adequate lung biopsy performed with bronchoscopic (forceps or cryobiopsy) or surgical approach (preferably, videoassisted thoracoscopic biopsy) can demonstrate the histopathologic pattern of lung injury (eg, nonspecific interstitial pneumonia [NSIP], organizing pneumonia [OP], or diffuse alveolar damage [DAD]) in patients suspected of having DRP. However, the features seen on lung biopsy are unlikely to confirm the diagnosis of DRP because these histopathologic patterns are nonspecific and can be seen with other causes, including infections. A biopsy may sometimes be useful to exclude recurrent malignancy, given that tumors can manifest as diffuse lung infiltration and mimic ILD.
Bronchoscopic cryobiopsy yields larger samples of lung tissue but is associated with a higher rate of bleeding and pneumothorax compared with forceps biopsy (39). Overall in-hospital mortality after surgical lung biopsy for ILD in the United States was found to be 6.4% in a recent analysis of a national data set (40). The in-hospital mortality rate was 1.7% for elective operations compared with 16.0% for nonelective operations. Possible need for surgical lung biopsy should be entertained early rather than late in the clinical course because severe respiratory dysfunction and dependence on mechanical ventilation increase the mortality rate associated with surgical lung biopsy (41).
More often, bronchoscopy with bronchoalveolar lavage is performed to exclude infections (including opportunistic, mycobacterial, and viral), alveolar hemorrhage, or metastatic and/ or lymphangitic spread (cancer cells). However, bronchoalveolar lavage fluid-derived differential cell count is often overlapping and nonspecific (27,28,42,43). Nonetheless, bronchoalveolar lavage may yield specimens diagnostic of infection (eg, Pneumocystis) or malignancy and bronchoalveolar lavage fluid-derived differential cell counts may provide diagnostic clues (eg, eosinophilia) as to the underlying pathologic process.
When Should Suspected Offending Drugs Be Stopped in Patients Suspected of Having DRP? Discontinuation of the suspected drug is advisable for patients with severe or progressive lung disease (eg, with worsening to grade 2 or 3 [Common Terminology Criteria for Adverse Events]) for which DRP is deemed a possible or likely cause of the clinical presentation, while additional studies are being performed for diagnostic clarification. Improvement following cessation of drug administration without glucocorticoid therapy would strongly support the diagnosis of DRP in the absence of other more likely explanations emerging from the diagnostic work-up (27,43).
It may be appropriate to closely monitor patients while continuing therapy when the lung injury is not severe or progressive, particularly asymptomatic patients with isolated radiologic changes (grade 1 pneumonitis), as discussed in the following sections regarding specific agents including mechanistic target of rapamycin, or mTOR, inhibitor and thirdgeneration EGFR-TKIs (32). It is important to acknowledge pulmonary embolism, pulmonary veno-occlusive disease, and other forms of pulmonary hypertension needs to be considered in this setting. The clinical suspicion for DRP arises from the temporal relationship between drug exposure and the onset of clinical presentation (27,28). In most patients, the diagnosis of DRP is unlikely to be made with certainty even after extensive clinical evaluation, including lung biopsy.
The increasing use of molecular targeting agents and ICIs has broadened the spectrum of lung toxicity encountered clinically, especially in patients with cancer. This is exemplified by immune-related adverse effects in patients treated with ICIs, which manifest as a wide array of organ toxicities, including pneumonitis (11,32,33). These toxicities are thought to be the result of general immunologic activation, including autoimmune response.
The severity of symptoms associated with DRP may range from mild to life-threatening with rapid progression to death. The Common Terminology Criteria for Adverse Events published by the National Cancer Institute of the National Institutes of Health provides standardized definitions for grading the severity of organ toxicity (34). The grades for pneumonitis include grade 1 (asymptomatic), grade 2 (symptomatic), grade 3 (severe symptoms), grade 4 (life-threatening respiratory compromise), and grade 5 (death related to adverse event).
Laboratory tests such as serologic testing and microbial cultures may help to establish infectious or other etiologies for pulmonary infiltrates but are not useful in specifically diagnosing DRP (12,(27)(28)(29)35,36). Similarly, pulmonary function testing commonly demonstrates a restrictive pattern (reduced forced vital capacity and/or total lung capacity) along with a reduced diffusion capacity, which is the typical pattern seen in ILD. When present, it is helpful in assessing the degree of pulmonary impairment, but it does not contribute to confirming the diagnosis of DRP (27)(28)(29).
When Should Chest CT Be Performed to Confirm the Diagnosis in Patients Suspected of Having DRP? In patients receiving drugs potentially causing pulmonary toxicity, chest CT (and particularly thin-section CT; section thickness of 2.0-2.5 mm or less) plays an important role in evaluating the appearance, the progression, and the resolution of pulmonary abnormalities (37). CT should be performed as early as possible when DRP is suspected and in the presence of a positive temporal relationship between drug exposure and symptom onset. CT may allow early detection of the DRP while it is still at a reversible stage, or it may help to identify findings indicating other etiologies that can explain the symptoms of the patients (38). CT is also essential to evaluate the presence of other common causes (eg, community-acquired or health care-associated pneumonia) for the nonspecific clinical manifestations of DRP.

Imaging Features of DRP
Most DRP is diagnosed on routine follow-up CT scans for monitoring in patients with cancer; specific protocol-based scans targeted to depict DRP are not obtained. In addition, given the unique nature (eg, nature of diagnostic exclusion, symptomless cases, and wide differential diagnoses, etc) of the entity, defining the specific protocol for CT study is difficult and impractical. Thinner-section (2.0-2.5 mm or less in section thickness) and contiguous CT scans are recommended usually with intravenous contrast agent injection. For thorough image analysis, not only transverse but also coronal reformatted images are needed. Follow-up chest CT is useful to assess the changes of DRP findings and response to DRP treatment (36,54). However, the details of follow-up scans including time intervals depend on the clinical context (severity of symptoms and clinical follow-up course, etc).
The CT features of DRP associated with systemic therapeutic agents should be systematically described, including distribution and patterns of parenchymal abnormalities and presence of individual features including ground-glass opacities, airspace consolidation, reticular opacities, centrilobular nodules, interlobular septal thickening, honeycombing, and traction bronchiectasis.
Each drug can be associated with multiple injury patterns at CT (Fig 1) (55), which are typically not specific. In most situations, clinicians rely on the temporal relationships between the administration of drugs and the onset of symptoms, along with the exclusion of other potential causes of lung injury, particularly infections and metastatic diseases (28).
What Are the CT Patterns of DRP? DRPs have various histologic patterns and diverse CT findings (31,56,57). Although CT and histologic patterns coincide in only half of patients with DRP (57), the CT pattern reflects the extent and distribution of lung abnormalities and helps to predict changes in terms of prognostication. Some of the commonly described patterns include interstitial pneumonia either as NSIP, OP, DAD, hypersensitivity pneumonitis (HP), and simple pulmonary eosinophilia (Table 1) (58)(59)(60)(61). Often, imaging demonstrates more than one pattern (Fig 1); in this case the dominant pattern is typically reported. Other drug-related lung diseases such as granulomatous pneumonitis, vasculitis, alveolar proteinosis, constrictive (obliterative) bronchiolitis, and veno-occlusive disease are uncommon (32,59) and demonstrate diverse CT features; thus, they are difficult to classify into one of the aforementioned common CT patterns. The DRP CT patterns are nonspecific for either drug reaction in general or the reaction to a particular drug. Consequently, the diagnosis of DRP is based on a combination of clinical, radiologic, and histologic (when necessary) findings in a patient who has received a drug known or suspected to cause the abnormalities.
Radiologic NSIP pattern.-NSIP pattern consists of patchy or diffuse areas of ground-glass opacity (58,59), typically with peripheral and lower lung zone predominance. With progression, evidence of fibrosis including reticulation, traction the life-threatening nature of the malignancies treated with the implicated drugs, the benefits of the therapy, and the uncertainties regarding the impact of medication discontinuation in this setting. Thus, it is appropriate for clinicians to discuss these issues at a multidisciplinary conference (see later section on Multidisciplinary Diagnosis of DRP) and with patients to reach a shared decision regarding preferred course of action. Additionally, major clinical guidelines based on the consensus of multidisciplinary and multiorganizational panels are available for specific entities such as ICI-related pneumonitis, and should be considered for patient treatment in specific clinical settings when relevant (44)(45)(46)(47).
Does Improvement with Glucocorticoid Therapy Confirm the Diagnosis of DRP? Clinical improvement subsequent to glucocorticoid therapy does not definitively confirm the diagnosis of DRP because other inflammatory processes that are not drug related (eg, radiation pneumonitis) may also respond to glucocorticoid therapy. In addition, improvement may merely be coincidental and due to a self-limited event (eg, aspiration pneumonia) with spontaneous recovery. Nonetheless, glucocorticoid therapy is commonly used in the management of patients suspected of having DRP, especially if the lung injury is severe or progressive, and response to glucocorticoid therapy would support a diagnosis of DRP (vs progression of underlying cancer, for example) in the absence of a better alternative explanation (27,45,48,49). Improvement with glucocorticoid therapy may also obviate invasive diagnostic maneuvers such as bronchoscopic or surgical lung biopsy.

Should Rechallenge with a Suspected Drug Be Performed to Confirm the Diagnosis of DRP?
It is rarely appropriate to rechallenge with the suspected drug to confirm the diagnosis, especially when the lung toxicity has been severe or if there were substantial residual abnormalities at chest imaging (45,50). An exception may be considered when lung toxicity has been mild and transient, particularly if alternative cancer therapies are unlikely to be effective.

What Prognostic Factors Should Be Considered in Patients with DRP?
Prognostic factors in patients with DRP include acute onset, severity of lung toxicity (eg, hypoxemia), response to treatment (eg, drug withdrawal), older age, current or prior smoking history, preexisting lung disease, other comorbidities, and the status of the underlying cancer (27,35). There are no scoring schemes currently available that integrate these factors into a predictive model. The prognosis associated with DRP also varies depending on the specific drug and the type of underlying cancer. There are emerging reports of beneficial effects from drug-related toxicities on tumor response to therapy, especially in those who are treated with newer molecular targeting cancer therapy or ICIs (51,52). This interesting possibility should be further investigated in a disease-specific and therapy-specific manner (53). chovascular in distribution (Fig 2) (62). The abnormalities are usually bilateral and symmetric, with predominant lower-lung bronchiectasis, and occasionally honeycombing are identified (Table 1). In some patients, fibrosis is predominantly peribron-

Pattern
Imaging Manifestations at Chest CT NSIP (see Fig 2) Starting from patchy areas of GGO, progression to irregular reticular opacities, architectural distortion, and traction bronchiectasis, with or without associated areas of consolidation; bilateral and symmetric, with predominant lower-lung involvement OP (see Figs 1c, 1d, 3, and 4) Multifocal patchy alveolar opacities typically with peribronchovascular and/or peripheral distribution; may demonstrate reversed halo sign HP (see Fig 5) Poorly defined small centrilobular nodules, bilateral GGO, large or lobular areas of decreased attenuation and vascularity (mosaic attenuation) DAD (AIP/ARDS pattern) (see Fig 6) Extensive bilateral areas of GGO and dependent airspace consolidation in exudative phase; traction bronchiectasis and decreased lung volumes in organizing and fibrotic phases Simple PEo (see Figs 1a, 1b, and 7) Nonsegmental consolidation or GGO, unilateral or bilateral; transient and migratory; spontaneous resolution within 4 weeks is common Note.-AIP = acute interstitial pneumonia, ARDS = acute respiratory distress syndrome, DAD = diffuse alveolar damage, GGO = groundglass opacity, HP = hypersensitivity pneumonitis, NSIP = nonspecific interstitial pneumonia, OP = organizing pneumonia, PEo = pulmonary eosinophilia.
Radiologic HP pattern.-HP pattern shows small, poorly defined centrilobular nodules with or without widespread areas of ground-glass opacity or lobular areas of decreased attenuation and vascularity (Fig 5). Radiologic HP pattern may occur after treatment with gefitinib or erlotinib (66), mTOR inhibitors, and ICIs (Table 2) (54,65).
Radiologic DAD pattern.-DAD pattern demonstrates extensive bilateral areas of ground-glass opacity and dependent airspace consolidation with traction bronchiectasis at chest CT with their proportion depending on disease phases (exudative, organizing, and fibrotic) (67). The extent of ground-glass opacity and traction bronchiectasis increases as the disease evolves (68). This pattern has been reported in patients treated with EGFR-TKIs, anaplastic lymphoma kinase inhibitors, and ICIs (Fig 6) (12,58,69). This radiologic pattern is often associated with serious clinical outcome from pneumonitis, thus requiring awareness of this pattern among radiologists (Table 2) (32).
Radiologic simple pulmonary eosinophilia pattern.-Simple pulmonary eosinophilia pattern demonstrates nonsegmental consolidation or ground-glass opacity that can be unilateral or bilateral. The lung abnormalities are usually transient and migratory, and the prognosis is excellent; spontaneous resolution within 4 weeks is common (Figs 1, 7) (70,71). The pulmonary eosinophilia pattern is seen in osimertinib therapy (72).

CT Characteristics Associated with Specific Classes of Cancer Therapy
With the recent rapid advances of cancer therapy, pneumonitis related to novel agents have been increasingly described, along with their CT patterns. Many studies, as described in the following sections, applied the concept of CT pattern-based approach similar to the one described above, indicating a widespread use and applicability of this approach. It should also be noted that the concept and approach to DRP continue to evolve, as more novel agents are translated into the clinical settings and provide newer sets of challenges for the diagnosis, monitoring, and treatment.

Molecular Target Agents
EGFR-TKI Therapy.-In 2003, four cases of severe DAD pattern in patients treated with gefitinib were reported; among 7 (74) reporting clinically significant effects of pneumonitis related to EGFR-TKIs. Serum proteomic markers and genetic polymorphisms have been studied as candidates to explain the higher incidence of pneumonitis in Japanese patients compared with others; however, no conclusive results have been obtained (73,(75)(76)(77). Genetic and environmental factors that contribute to the development of EGFR-TKI pneumonitis remain to be understood. NSIP (Fig 2), OP (Fig 1), DAD (Fig 6), and HP ( Fig 5) patterns have been reported with EGRF-TKIs. Poor prognosis is expected when there is a short interval between the initiation of the targeting therapy and the onset of pneumonitis, when the CT findings are represented by a DAD pattern, and when there is preexisting ILD (58). Severe and potentially fatal four patients, two recovered with steroids and two died due to the DRP (9). In a recent report of a meta-analysis (73) of 153 trials worldwide including 15 713 patients with NSCLC and EGFR-TKI therapy, the overall incidence of DRP was 1.12% for all grades, 0.61% for high-grade pneumonitis, and 0.20% for grade 5 pneumonitis. When the incidence of pneumonitis was compared among the multiple factors including EGFR-TKI agents, treatment lines, EGFR mutation status, trial phases, and countries in the meta-analyses, significantly higher incidence rates were noted among Japanese studies compared with non-Japanese studies for all grades (4.77% vs 0.55%; P , .001), high grade (2.49% vs 0.37%; P , .001), and grade 5 pneumonitis (1.00% vs 0.18%; P , .001) ( Table 2). These findings provide further support to a previous Japanese study Note.-Modified from reference 24. Data in parentheses are 95% CIs. Drug-related pneumonitis from mechanistic target of rapamycin inhibitors, CD20 antibodies, and ipilimumab were not tabulated owing to lack of robust meta-analysis data. AIP = acute interstitial pneumonia, ALK = anaplastic lymphoma kinase, ARDS = acute respiratory distress syndrome, DAD = diffuse alveolar damage, EGFR = epidermal growth factor receptor mutation, HP = hypersensitivity pneumonitis, NSCLC = non-small cell lung cancer, NSIP = nonspecific interstitial pneumonia, OP = organizing pneumonia, PD-1 = programmed cell protein death 1, PD-L1 = programmed death ligand 1, PeO = pulmonary eosinophilia. Source.-References 54, 58, 73, 84, 85, 89, 90, 120. * Incidence rates are meta-analyses of trials of NSCLC treated with single-agent therapy. † Incidence is among patients treated with EGFR inhibitors without prior exposure to EGFR-directed therapy. ‡ Data include patients who received osimertinib after previous treatment with conventional EGFR inhibitors. Overall incidence of pneumonitis was 4% in a recent phase 3 first-line treatment of osimertinib for EGFR-mutant NSCLC. § In addition to these common patterns, "pulmonary edema-like shadows" characterized by bilateral ground-glass appearance, thickening of the interlobular septa and the bronchovascular bundles distributed predominantly in the side of the pulmonary hilum, and occasional bilateral pleural effusion have been described in ALK-related pneumonitis. ║ Incidence rates are based on the meta-analyses of PD-1 inhibitor trials for melanoma, NSCLC, and renal cell carcinomas. # Incidence rates are based on the meta-analyses of combination therapy regimens of PD-1 inhibitor, combined with ipilimumab or peptide vaccines, for patients with melanoma. ** Incidence rates are based on the meta-analyses of single-agent PD-L1 inhibitor trials for NSCLC. † † The study addressed only grade 3-4 pneumonitis.  (53). In 66 patients with advanced neuroendocrine tumors who were treated with everolimus, DRP was reported in 14 (21%) patients (OP pattern in eight, NSIP pattern in five, and HP pattern in one) (Fig 4) (65). In 40 patients with Waldenstrom macroglobulinemia being treated with everolimus, 23 (58%) patients developed DRP, with a radiologic OP pattern in 16 and NSIP pattern in seven (82).
According to the management guideline by Albiges et al (83), asymptomatic patients with mTOR pneumonitis and radiologic changes only (grade 1) may continue mTOR inhibitor therapy without dose adjustment at the treating physician's discretion. However, patients should be informed of any signs of worsening to look out for, which would require contacting their physician. pneumonitis in patients treated with ICIs plus EGFR-TKIs has been recently reported (56,78).
Emerging observations indicate that a milder form of lung reaction to EGFR-TKIs may manifest at imaging only without clinical symptoms, especially in the setting of the newer EGFR-TKI treatments. A novel type of drug-related pulmonary phenomenon called transient asymptomatic pulmonary opacities has been described in up to 20% of patients with NSCLC treated with the third-generation EGFR-TKI osimertinib (Fig 8)  (72,79). Transient asymptomatic pulmonary opacities are described as localized pulmonary opacities mostly with radiologic simple pulmonary eosinophilia pattern, which resolves without any treatment during continued osimertinib therapy with a median 6-week duration (Figs 1, 8). Interestingly, patients who developed this apparent grade 1 pneumonitis had longer progression-free survival and overall survival compared with patients without transient asymptomatic pulmonary opacities, indicating the potential association between drug-related phenomena and treatment benefits. mTOR inhibitors.-DRP is frequently seen in patients with mTOR inhibitors, including temsirolimus and everolimus. Temsirolimus has been approved for treatment of renal cell

ICI Therapy
The U.S. Food and Drug Administration has approved agents including ipilimumab (cytotoxic T-lymphocyte-associated protein 4 inhibitor), nivolumab, and pembrolizumab (programmed cell death protein 1 [PD-1] inhibitors), as well as atezolizumab and durvalumab (programmed death ligand 1 [PD-L1] inhibitors) to treat different types of advanced cancer (87). In this setting, ICI therapy is associated with a variety of immune-related adverse effects that can affect any organ (6,56). The initial reports have described a spectrum of radiologic patterns of interstitial pneumonias and clinical courses ( Table 2) (12,35,36,88).
Anaplastic lymphoma kinase inhibitors.-Severe acute pneumonitis in patients receiving crizotinib therapy for advanced NSCLC has been reported (69). In the recent meta-analysis (84) of 18 trials with 2261 patients with anaplastic lymphoma kinase inhibitor monotherapy and advanced NSCLC, the overall incidence of pneumonitis was 2.14% for all grades, 1.33% for high-grade pneumonitis (grade 3 or above), and 0.22% for grade 5 pneumonitis. Similar to the EGFR-TKI study, Japanese cohorts showed a higher incidence of anaplastic lymphoma kinase-inhibitor pneumonitis for all grades (6.25% vs 1.14%; P , .001) and grade 3 and above pneumonitis (3.31% vs 0.39%; P , .001), compared with non-Japanese cohorts from multiple countries other than Japan (Table 2) (84). In postmarketing surveillance of crizotinib therapy in Japan, the incidence of pneumonitis associated with crizotinib therapy was 5.8% for all grades, and 3.5% for grade 3 or greater pneumonitis. In 27% of patients with pneumonitis, CT findings were suggestive of the presence of DAD. Age 55 years or older, Eastern Cooperative Oncology Group performance status between 2 and 4, smoking history, previous or concomitant ILD, and comorbid pleural effusion were noted as significant risk factors for crizotinib-related pneumonitis (85). monotherapy. In another meta-analysis (90) of 19 clinical trials of PD-1 inhibitors and PD-L1 inhibitors as single-agent therapy in NSCLC, the incidence was higher in patients treated with PD-1 inhibitors compared with those treated with PD-L1 inhibitors (3.6% vs 1.3%, respectively; P = .001), providing valuable insight for optimal clinical selection of these agents given the overlapping approved indications of PD-1 and PD-L1 inhibitors. A subanalysis of patients with NSCLC treated with pembrolizumab in the phase I KEYNOTE-001 trial demonstrated that the overall incidence of In a meta-analysis (89) including 4496 patients from 20 single-tumor-type trials of PD-1 inhibitor including 12 melanoma studies, five NSCLC studies, and three renal cell carcinoma studies, the overall incidence of pneumonitis during PD-1 inhibitor monotherapy was 2.7% (95% CI: 1.9, 3.6) for all grades and 0.8% (95% CI: 0.4, 1.2) for grade 3 or higher pneumonitis. The incidence of PD-1-related pneumonitis was higher in patients with NSCLC or renal cell carcinoma compared with that in patients with melanoma, and during combination therapy compared with  of lung injury or in whom the differential diagnosis raises the consideration of markedly different therapeutic strategies (eg, drug toxicity vs infection or malignancy).

What Histologic Characteristics Should Be Documented in Lung Biopsies Performed for DRP?
Lung biopsies should be evaluated for patterns of interstitial pneumonia by using criteria within the American Thoracic Society/ European Respiratory Society International Multidisciplinary Classification of the Idiopathic Interstitial Pneumonias (94,95) including cellular and fibrotic NSIP, usual interstitial pneumonia, OP (including acute fibrinous subtype), lymphoid interstitial pneumonia and DAD, as well as bronchocentric inflammatory changes (including hypersensitivity pneumonia) and noncaseating granulomas (96). Diffuse malignant infiltration, mimicking or coexisting with ILD, should be ruled out. In addition, depending on the morphologic features and the clinical setting, infectious agents such as bacteria, fungi, mycobacteria, or viral agents should be searched for by using special stains where indicated.

What Histologic Features Are Most Suggestive of DRP?
Although any of the above histologic patterns can be seen in DRP, there is a more frequent overlap of patterns, coexistent tissue eosinophilia, chronic interstitial inflammation, lymphoid aggregates, and pleuritis compared with idiopathic cases showing similar histologic patterns. However, these same features are not specific because they are also seen in connective tissue disease-related lung disease.
There are limited published data on the pathologic features of pulmonary toxicity in ICIs and targeting molecular therapies; hence, most of the cases are diagnosed based on clinical and CT features only. Nonetheless, the main pathologic features described include cellular and/or fibrosing interstitial pneumonia, OP, HP, DAD, and pulmonary eosinophilia (35,54).

What Information Is Available from Bronchoalveolar Lavage Fluid Analysis?
Infectious organisms can be identified with bronchoalveolar lavage fluid cultures (27). In 12 (46%) of 26 patients with pneumonitis was 3.8%. A higher incidence was noted in patients with a history of asthma or chronic obstructive pulmonary disease (5.3%) and in those with a history of thoracic radiation (6.0%) (6,91). A retrospective study enrolling 1826 patients with cancer reported 64 (3.5%) cases of ICI-related pneumonitis, which more commonly occurred in men and former or current smokers, with a median age of 59 years. In this series, 66% of patients with pneumonitis had grade 2 or 3, 9% had grade 4, and 9% had grade 5 (fatal) pneumonitis. An earlier onset was noted in lung cancer versus melanoma (median of 2.1 months vs 5.2 months; P = .02). OP (23%) was the most common pattern followed by HP pattern (16%) (87,92). Moreover, in a recent meta-analysis of fatal toxicities related to ICIs, DRP was identified as the most common toxicity leading to PD-1/PD-L1-related mortality, accounting for 35% of all deaths (88,93). In a study of 20 patients with DRP among 170 patients with melanoma, lung cancer, and lymphoma treated with PD-1 inhibitors, OP pattern (Fig 3) was found in 13 patients and was the most common pattern, followed by NSIP pattern in three, HP pattern in two, and DAD pattern (acute respiratory distress syndrome/acute respiratory distress syndrome pattern in the study referring to the American Thoracic Society/European Respiratory Society classification of idiopathic interstitial pneumonias) in two patients (54). The CT patterns were associated with the toxicity grades of pneumonitis as defined by Common Terminology Criteria for Adverse Events; DAD pattern had the highest grades, followed by OP pattern, whereas NSIP and HP patterns had lower grades, indicating the utility of CT pattern-based approach in assessing the severity of ICI-related pneumonitis (54). Seventeen patients received corticosteroid therapy, and three also received infliximab treatment. Seven patients were retreated, of whom two developed DRPs again. One patient demonstrated pneumonitis flare-up on tapering of corticosteroid intake without retreatment with ICI or any other agents, further indicating the complex nature of the entity and importance of imaging follow-up of these patients (54).

Pathologic Analysis
Lung biopsy may be indicated in patients in whom the clinical and radiologic picture do not clearly point to a specific pattern  shown to be effective in other disciplines (99). Multidisciplinary diagnosis is particularly important in patients suspected of having DRP because there is no individual feature that is required or sufficient for the diagnosis of DRP. The multidisciplinary diagnosis approach typically involves clinicians, radiologists, and pathologists (if biopsy is performed) and can be used in clinical settings and trials with centralized review of adverse events. Because DRP is often observed in more acute or subacute clinical settings, the actual consultations among subspecialties may happen as informal and formal communication by using telephone or virtual conference. In more chronic or difficult cases, it may be discussed formally at a multidisciplinary diagnosis conference. It is important that multidisciplinary discussion shall happen in the clinical context of the need for clinical management of the patients suspected of having DRP.

Description of Central Review for the Multidisciplinary Diagnosis of DRP
To determine the accurate incidence of DRP in clinical trials and postmarketing surveillance, the cases diagnosed by each physician in primary investigation site should be evaluated by using a process of central review (100,101) to achieve uniform criteria through accurate and consistent data. The strategy at the time of the review should be based on a multidisciplinary diagnosis approach, involving a multidisciplinary team consisting of at least one chest physician, one oncologist, one chest radiologist, and (if a biopsy is available) one pathologist. Moreover, it is essential to use a mutually agreed diagnostic checklist (refer to Appendix E2 [online] for record of multidisciplinary discussion and Appendix E3 [online] for objective evaluation of chest CT and ILD) consistently throughout an individual study or cohort. At first, each radiologist and chest physician should independently evaluate the case, followed by the subsequent multidisciplinary discussion among the experts to reach a consensus.

Management
Pharmacovigilance, or drug safety monitoring, plays an important role in identifying, understanding, and preventing adverse drug reactions. The World Health Organization Program for International Drug Monitoring (VigiAccess) provides an international forum for collaboration in pharmacovigilance, collecting data from real-world settings. All drug-related adverse effects should be declared to the phar-everolimus treatment and initial diagnosis of DRP, bronchoalveolar lavage fluid cultures enabled a diagnosis of Pneumocystis jirovecii pneumonia (97). Cell count of bronchoalveolar lavage fluid in DRP may disclose raised lymphocyte, neutrophil, and eosinophil numbers. Particularly in pulmonary eosinophilia pattern, cell count helps to make a diagnosis of the disease by documenting eosinophilia. However, the cell count is not specific, because similar results could be seen in other inflammatory and infectious conditions (27,28,42,43).

Proposal for Diagnostic Criteria
Camus et al (98) proposed the following diagnostic criteria for DRP: (a) exposure to the causative drug, (b) development of pulmonary infiltrates, (c) meticulous exclusion of all other possible causes, (d) dechallenge producing measurable improvement in symptoms and imaging, and (e) rechallenge causing worsening. However, some patients do not have improvement with dechallenge, and rechallenge is often impossible in many clinical settings, thus making these criteria impractical in many patients.
Therefore, we propose the following criteria: (a) newly identified pulmonary parenchymal opacities at CT or chest radiography, commonly in a bilateral nonsegmental distribution; (b) temporal association of presentation with the initiation of a systemic therapeutic agent; and (c) exclusion of other likely causes (

Introduction of Multidisciplinary Diagnosis
The process of multidisciplinary diagnosis is by means of interactive multidisciplinary discussion, an approach that has been  without ILD. Furthermore, greater CT extent of preexisting ILD portended higher risk of fatal outcome.

Does Glucocorticoid Therapy Improve Clinical Outcome in Patients with DRP?
Glucocorticoid therapy is often used in patients with DRP to ameliorate and to expedite the recovery of lung injury (27,45,48). This strategy is commonly used when DRP is moderate to severe and of acute or fulminant onset. However, this practice is based on retrospective studies and expert opinion because no clinical trial has been performed to prove the efficacy of glucocorticoid therapy in the treatment of patients with DRP (113).

Conclusion
This position paper of the Fleischner Society summarizes simplified diagnostic criteria, CT pattern approach, and management recommendation of drug-related pneumonitis (DRP) in the emerging era of molecular targeting agents and cancer immunotherapy, by using a multidisciplinary approach. The diagnosis and management of DRP will continue to evolve with the advancement of treatments, and a radiologic pattern approach with multidisciplinary diagnosis will remain crucially important for the optimal treatment of the patients.  macovigilance program, which is the cornerstone of the alert system. With clinicians' reports on adverse drug reactions, the incorporation of clinical and pharmacologic information could help to avoid the unnecessary exposure to adverse drug reactions (102,103). In general, delayed diagnosis of DRP is associated with higher severity of lung injury and less reversibility, resulting in residual lung damage (ie, fibrosis) (27,48). Thus, early diagnosis and cessation (except some drugs; refer to previous sections regarding specific agents) of the offending drug intake promote optimal outcomes in patients with DRP. In addition, glucocorticoids are commonly administered to facilitate the resolution of lung injury, particularly for those severely affected (National Cancer Institute grade 3-4 pneumonitis) as assessed by symptoms, gas exchange derangements, and radiologic abnormalities (27,35,45,48). These patients usually require hospitalization for their initial treatment and monitoring (35,45). Supportive measures, including supplemental oxygen and noninvasive or invasive mechanical ventilator support, may be needed.

Disclosures of Conflicts
For the management of ICI-related pneumonitis, the National Comprehensive Cancer Network, American Society of Clinical Oncology, Society for Immunotherapy of Cancer, and European Society for Medical Oncology guidelines (45)(46)(47)104,105) recommend discontinuing ICI therapy for any grade of pneumonitis and recommend treating grade 2 pneumonitis with corticosteroids. Permanent discontinuation of ICIs is suggested for grade 3-4 pneumonitis by all guidelines. For those without improvement on corticosteroids after 48 hours, infliximab, mycophenolate mofetil, or intravenous immunoglobulin may be used (45)(46)(47)104,105).

Is Preexisting ILD a Risk Factor for DRP and Does Preexisting ILD Lead to Worse Outcomes in Patients with DRP?
Preexisting ILD is a risk factor for the development of DRP (27,28,48). Multiple studies on the incidence of lung toxicity associated with antineoplastic agents have demonstrated that preexisting ILD is associated with a higher likelihood of DRP, sometimes described as an acute exacerbation of preexisting ILD (28,74,80,(106)(107)(108)(109)(110)(111)(112). For example, the odds ratio for developing DRP in patients with lung cancer treated with chemotherapy or gefitinib ranges from 4.8 to 25.3 (depending on the severity of ILD) in patients with preexisting ILD compared with those without preexisting ILD (74). A recent systematic review (113) on DRP found that preexisting ILD is an independent risk factor for DRP with a wide spectrum of therapeutic agents. However, some uncertainty remains on whether preexisting ILD is a risk factor for the development of ICI-associated pneumonitis in particular, largely due to the exclusion of patients with preexisting ILDs from clinical trials of ICIs. Recent studies have suggested preexisting fibrotic changes at CT are associated with an increased risk of anti-PD-1-related pneumonitis in patients with NSCLC (114,115).
Several studies demonstrated worse outcome related to DRP among patients with preexisting ILD compared with those without (74,109,110,112,(116)(117)(118)(119). For example, Kudoh et al (110) reported the odds ratio for fatal outcomes from DRP related to chemotherapy or gefitinib therapy to be 2.27 for patients with preexisting ILD compared with those