Global Disproportionality Analysis of Adverse Event Reports on Interstitial Lung Diseases for Cancer‐Targeted Therapies, 1968–2024

ABSTRACT

Background and Objectives

The increasing complexity of interstitial lung disease (ILD) related to cancer‐targeted monoclonal antibodies (mAbs) has emerged as a significant clinical concern. Thus, this study aimed to investigate reporting signals of four ILD subtypes detected with cancer‐targeted mAbs.

Methods

This global pharmacovigilance study conducted disproportionality analyses to detect signals of ILD subtypes reported with cancer‐targeted mAbs. ILD subtypes were classified into eight categories according to previous studies, of which four with a low number of reports were excluded. Five cancer‐targeted monoclonal antibodies (VEGF/VEGFR, CD20, PD‐1/PD‐L1, HER2, and EGFR inhibitors) were included according to ATC code (L01F). Reporting signals were evaluated using reporting odds ratio (ROR) with 95% CI and information component (IC) with IC₀₂₅.

Results

Interstitial pneumonitis and pulmonary fibrosis showed significant disproportionate reporting signals across all drugs. Notably, interstitial pneumonitis showed significant signals with EGFR inhibitors (ROR, 47.46 [95% CI, 44.67–50.42]; IC, 5.47 [IC_0.25, 5.37]) and PD‐1/PD‐L1 inhibitors (45.20 [43.49–46.97]; 5.36 [5.30]), while pulmonary fibrosis showed higher signals with CD20 inhibitors (12.71 [11.30–14.31]; 3.61 [3.42]). Organising pneumonia exhibited significant signals with PD‐1/PD‐L1 inhibitors (ROR, 44.48 [95% CI, 39.05–50.67]; IC, 5.26 [IC_0.25, 5.04]), followed by CD20 (17.45 [13.88–21.94]; 3.95 [3.57]), and HER2 (15.95 [11.88–21.40]; 3.76 [3.26]), with the strongest signal observed with PD‐1/PD‐L1 inhibitors. Eosinophilic pneumonia showed significant signals with all drugs except CD20 inhibitors.

Conclusions

Although causal inference cannot be drawn, this global disproportionality study highlights reporting signals between cancer‐targeted mAbs and ILD subtypes, underscoring the importance of strengthening adverse event reporting systems before and after mAb administration.

Short abstract

This global disproportionality analysis revealed that cancer‐targeted monoclonal antibodies (mAbs) show distinct disproportionate reporting signals for specific interstitial lung disease subtypes. Interstitial pneumonitis and pulmonary fibrosis were most strongly linked to EGFR, PD‐1/PD‐L1, and CD20 inhibitors. These findings highlight the importance of subtype‐specific monitoring in patients receiving mAb therapies.

See related editorial

1. Introduction

Interstitial lung disease (ILD) is a complex and heterogeneous condition caused by adverse events of various therapeutic agents [1]. This group of disorders has become increasingly significant with the rapid introduction of novel medications, particularly as advancements in cancer therapies continue to accelerate [2]. Among these, the introduction of monoclonal antibodies (mAbs) as cancer treatments has notably contributed to an increase in ILD adverse events [3]. While the prevalence and mortality rates of other respiratory diseases, such as COPD and asthma, declined from 1990 to 2019, ILD showed a 9.4% increase in prevalence and a 23.4% rise in mortality [4]. During this period, the US Food and Drug Administration approved several classes of cancer‐targeted monoclonal antibodies, beginning with clusters of differentiation 20 (CD20) inhibitors (e.g., rituximab) in 1997, human epidermal growth factor receptor 2 (HER2) inhibitors (e.g., trastuzumab) in 1998, epidermal growth factor receptor (EGFR) inhibitors (e.g., erlotinib, cetuximab) in 2004, vascular endothelial growth factor (VEGF/VEGFR) inhibitors (e.g., bevacizumab, pegaptanib) in 2004, and programmed cell death protein 1/death ligand 1 (PD‐1/PD‐L1) inhibitors (e.g., pembrolizumab) in 2014 [5, 6, 7, 8, 9]. Thus, it is crucial to systematically analyse the influence of specific drugs on ILD, classified by its subtypes.

Despite its importance, comprehensive studies investigating the differential effects of drugs on various ILD subtypes remain limited. Most studies on ILD have focused on the impact of individual drugs or drug classes on ILD collectively, with limited attention to analyses that distinguish between specific subtypes [10]. Additionally, although some studies have aggregated findings from prior research to evaluate the potential links between related drugs and the development of ILD, many were limited by a small number of ILD cases [11] or by low reliability due to the poor quality of the included studies [12].

Therefore, this global disproportionality study aimed to conduct signal detection analyses of cancer‐targeted mAbs and the development of four subtypes of ILD using the global adverse drug reaction (ADR) database. By providing a detailed analysis that considers the unique characteristics of each ILD subtype, this research seeks to advance understanding of the interplay between mAbs and ILD. Through this study, we endeavour to support healthcare professionals in enhancing ILD surveillance and monitoring reported events following drug administration.

2. Methods

2.1. Data Source

This study conducted signal detection analyses of cancer‐targeted mAbs in relation to ILD subtypes using a global dataset monitoring ADR, as part of the international drug adverse events monitoring program overseen by the World Health Organization (WHO) [13]. Additionally, this dataset encompasses data from over 140 countries and contains more than 35 million individual case safety reports (ICSRs). These reports include patient information, adverse reaction details, and drug information, all standardised according to the MedDRA terminology system (version 26.0) [14]. ICSRs are voluntarily reported by healthcare professionals, pharmaceutical companies, and patients, and the submitted data are regularly evaluated and analysed to ensure data integrity and reliability based on predefined criteria [15]. Therefore, the classification of ILD subtypes in this study was derived from MedDRA coding within the submitted ICSRs rather than through central adjudication.

2.2. Selection of Cases

This study conducted signal detection analyses of four ILD subtypes in relation to monoclonal antibodies from 1968 to 2024. Cancer‐targeted mAb classified under ATC code L01F were selected as the subjects of analysis [3, 16, 17]. However, drugs in the L01FB and L01FC subgroups were excluded because the number of reports related to ILD adverse events in the database was only 6 and 37, respectively, making it difficult to secure statistical significance: (1) CD20 inhibitors, (2) HER2 inhibitors, (3) EGFR inhibitors, (4) PD‐1/PD‐L1 inhibitors, (5) VEGF/VEGFR inhibitors. Additionally, ILD subtypes reported with adverse events were classified into the following five categories based on previous literature [18]: (1) pleuroparenchymal fibroelastosis, (2) interstitial pneumonia, (3) eosinophilic pneumonia, (4) organising pneumonia, (5) pulmonary fibrosis. In this study, idiopathic and cryptogenic cases, which refer to diseases of unknown aetiology, were excluded because it is difficult to reasonably conclude that these diseases were caused by drug exposure. Additionally, the analysis excluded pleuroparenchymal fibroelastosis because fewer than 20 reported cases were reported. Adverse events were collected using MedDRA version 26.0, focusing on the preferred terms listed in Table S1. Only cases identified as involving “suspected” drugs, according to WHO causality assessment guidelines, were included for signal detection analyses with ILD subtypes.

2.3. Data Collection

This study systematically documented reported ILD subtypes potentially related to drug exposure through a thorough investigation utilising a global pharmacovigilance dataset. The dataset used in this study was categorised into three main domains. First, patient demographics were classified by age groups, sex, and reporting regions. Second, ADRs were categorised by reporting periods, time to onset (TTO), and outcomes. Third, drug information was classified based on drug type and suspected drug [19].

2.4. Statistical Analysis

The dataset used in this study was analysed by dividing it into “reported” and “non‐reported” groups. Based on this classification, the study conducted signal detection analyses to identify disproportionate reporting of ILD subtypes in relation to each drug. For variables such as clinical outcome, sex, and drug exposure, missing values were retained and categorised as “unknown” in the dataset. To this end, the study employed two pharmacovigilance metrics: the reporting odds ratio (ROR) and the information component (IC). The ROR quantifies the extent to which reports of specific ILD subtypes were disproportionately observed with a given drug compared to all other drugs in the database. A disproportionate signal was considered present if the lower bound of the 95% confidence interval (CI) exceeded 1.00. The IC was derived using a Bayesian statistical approach to compare the reported and non‐reported groups. In keeping with standard IC analysis, an event was regarded as disproportionately over‐reported if the lower endpoint of the 95% credibility interval for the IC was positive (IC_0.25 > 0.00). Additionally, statistical significance was determined based on a two‐sided p‐value of less than 0.05. All analyses in this study were performed using SAS software (version 9.4; SAS Institute, Cary, NC, USA).

3. Results

3.1. Overall Analysis

After analysing reports from 1968 to 2024, we identified four ILD subtypes potentially reported in relation to five different cancer‐targeted mAbs, including a total of 7010 reports in the analysis (Figure 1). The four ILD subtypes linked through disproportionate reporting signals to cancer‐targeted monoclonal antibodies showed a sharp increase in cases during 2010 and 2020. In the Western Pacific region, interstitial pneumonitis and eosinophilic pneumonia were most frequently reported, with 4892 cases (82.75%) and 50 cases (73.53%), respectively. Organising pneumonia was most frequently reported in the Eastern Mediterranean region, with 179 cases (48.38%). Pulmonary fibrosis was predominantly reported in the region of the Americas (250 cases, 37.88%) and the European region (292 cases, 44.24%). Among interstitial pneumonitis and organising pneumonia, reports related to PD‐1/PD‐L1 inhibitors were the most frequently reported, with 2830 cases (47.87%) and 244 cases (65.95%), respectively. For pulmonary fibrosis, adverse event reports related to CD20 inhibitors were the most common, with 281 cases (42.58%), while VEGF/VEGFR inhibitors accounted for the highest number of adverse events for eosinophilic pneumonia, with 26 cases (38.24%). For clinical features, the proportion of recovered cases was high in interstitial pneumonitis (recovered/recovering: 3449 [58.34%]) and organising pneumonia (recovered/recovering: 227 [61.35%]), whereas pulmonary fibrosis (not recovered: 154 [23.33%]; fatal: 100 [15.15%]) and eosinophilic pneumonia (not recovered: 11 [16.18%]; fatal: 22 [32.35%]) showed a relatively higher proportion of not recovered and fatal cases (Figure 2 and Table 1).

FIGURE 1.

Flowchart of the selection process for drug‐associated ILD cases. ILD, interstitial lung disease.

FIGURE 2.

Proportional distribution of adverse event reports for four ILD subtypes associated with monoclonal antibodies. CD20, clusters of differentiation 20; EGFR, epidermal growth factor receptor; HER2, human epidermal growth factor receptor 2; ILD, interstitial lung disease; PD‐1/PD‐L1, programmed cell death protein receptor; VEGF/VEGFR, vascular endothelial growth factor.

TABLE 1.

Baseline characteristics of drug‐associated adverse reports for four ILD subtypes in the global pharmacovigilance database from 1968 to 2024.

Variables		Subtypes of ILD adverse events
		Interstitial pneumonitis, n (%)	Organising pneumonia, n (%)	Pulmonary fibrosis, n (%)	Eosinophilic pneumonia, n (%)
Region reporting	African Region	3 (0.05)	0 (0.00)	3 (0.45)	0 (0.00)
	Region of the Americas	64 (1.08)	98 (26.49)	250 (37.88)	5 (7.35)
	South‐East Asia Region	3 (0.05)	3 (0.81)	6 (0.91)	0 (0.00)
	European Region	950 (16.07)	90 (24.32)	292 (44.24)	13 (19.12)
	Eastern Mediterranean Region	0 (0.00)	179 (48.38)	27 (4.09)	0 (0.00)
	Western Pacific Region	4892 (82.75)	82 (12.42)	82 (12.42)	50 (73.53)
Reporting year	1968–1999	0 (0.00)	0 (0.00)	0 (0.00)	0 (0.00)
	2000–2009	32 (0.54)	1 (0.27)	41 (6.21)	0 (0.00)
	2010–2019	5227 (88.41)	269 (72.70)	338 (51.21)	61 (89.71)
	2020–2024	653 (11.05)	100 (27.03)	281 (42.58)	7 (10.29)
Reporter qualification	Health Professional	5783 (97.82)	364 (98.38)	544 (82.42)	68 (100.00)
	Non‐Health Professional	4 (0.07)	3 (0.81)	77 (11.67)	0 (0.00)
	Unknown	125 (2.11)	3 (0.81)	39 (5.91)	0 (0.00)
Sex	Male	3895 (65.88)	226 (61.08)	367 (55.61)	51 (75.00)
	Female	1919 (32.46)	137 (37.03)	280 (42.42)	17 (25.00)
	Unknown	98 (1.66)	7 (1.89)	13 (1.97)	0 (0.00)
Age	18–44 years	206 (3.48)	46 (12.43)	30 (4.55)	5 (7.35)
	45–64 years	1719 (29.08)	122 (32.97)	231 (35.00)	27 (39.71)
	≥ 65 years	3987 (67.44)	202 (54.59)	399 (60.45)	36 (52.94)
Delay (TTO), days	Median (Q1–Q3)	1 (1–1)	1 (1–1)	1 (1–29)	1 (1–1)
Drug class	CD20 inhibitors	407 (6.88)	75 (20.27)	281 (42.58)	1 (1.47)
	HER2 inhibitors	888 (15.02)	45 (12.16)	85 (12.88)	8 (11.76)
	EGFR inhibitors	1096 (18.54)	1 (0.27)	35 (5.30)	13 (19.12)
	PD‐1/PD‐L1 inhibitors	2830 (47.87)	244 (65.95)	233 (35.30)	20 (29.41)
	VEGF/VEGFR inhibitors	691 (11.69)	5 (1.35)	26 (3.94)	26 (38.24)
Fatal outcomes a	Recovered/recovering	3449 (58.34)	227 (61.35)	115 (17.42)	30 (44.12)
	Not recovered	855 (14.46)	40 (10.81)	154 (23.33)	11 (16.18)
	Fatal	995 (16.83)	22 (5.95)	100 (15.15)	22 (32.35)
	Death	0 (0.00)	0 (0.00)	2 (0.30)	0 (0.00)
	Unknown	613 (10.37)	81 (21.89)	289 (43.79)	5 (7.35)
Single drug suspected		5912 (100.00)	370 (100.00)	660 (100.00)	68 (100.00)

Abbreviations: CD20, clusters of differentiation 20; EGFR, epidermal growth factor receptor; HER2, human epidermal growth factor receptor 2; ILD, interstitial lung disease; IQR, interquartile range; PD‐1/PD‐L1, programmed cell death protein 1/death ligand 1; TTO, time to onset; VEGF/VEGFR, vascular endothelial growth factor.

^a Recovered/recovering indicates patients who have recovered or are recovering at the time of reporting; not recovered refers to those who had not recovered; fatal denotes life‐threatening or potentially fatal reactions not necessarily resulting in death; death indicates cases where death was explicitly reported; unknown refers to cases with no outcome information available.

3.2. Analysis of the Disproportionate Reporting Signals of ILD Subtypes Detected With Drugs

Overall, the use of the five selected drugs showed significant disproportionate reporting signals for four ILD subtypes (interstitial pneumonitis: ROR, 33.06 [32.15–34.00]; IC, 4.81 [IC_0.25, 4.76]; organising pneumonia: ROR, 22.60 [20.27–25.20]; IC, 4.31 [4.13]; pulmonary fibrosis: ROR, 7.26 [6.72–7.86]; IC, 2.81 [2.68]; eosinophilic pneumonia: ROR, 4.65 [3.66–5.92]; IC, 2.15 [1.75]). Even when analysed by sex, both males and females exhibited significant signals, with males (interstitial pneumonitis: 3895; organising pneumonia: 226; pulmonary fibrosis: 367; eosinophilic pneumonia: 51) showing more reports than females (interstitial pneumonitis: 1919 cases; organising pneumonia: 137; pulmonary fibrosis: 280; eosinophilic pneumonia: 17). In interstitial pneumonitis, stronger signals were observed in females (female: ROR, 33.52 [95% CI, 31.95–35.17]; IC, 4.86 [IC_0.25, 4.78]; male: ROR, 27.90 [26.95–28.88]; IC, 4.53 [4.48]). In addition, age‐stratified analysis showed that all drugs showed disproportionate reporting signals of interstitial pneumonitis and organising pneumonia across all age groups. All drugs, except VEGF/VEGFR inhibitors, showed signals of disproportionate reporting for pulmonary fibrosis in individuals aged 65 years and older (CD20 inhibitors: IC, 3.20 [IC_0.25, 2.94]; HER2 inhibitors: 2.47 [1.95]; EGFR inhibitors: 1.37 [0.69]; PD‐1/PD‐L1 inhibitors: 2.35 [2.09]) (Table S2 and Table S3).

All five selected drugs showed disproportionate reporting signals for interstitial pneumonitis and pulmonary fibrosis. Notably, patients receiving EGFR inhibitors had a significantly higher signal risk of developing interstitial pneumonitis (ROR, 47.46 [95% CI, 44.67–50.42]; IC, 5.47 [IC_0.25, 5.37]), as did those receiving PD‐1/PD‐L1 inhibitors (ROR, 45.20 [43.49–46.97]; IC, 5.36 [5.30]). In addition, signals for pulmonary fibrosis were most pronounced in reports linked to CD20 inhibitors (ROR, 12.71 [11.30–14.31]; IC, 3.61 [3.42]). The signal risk of developing organising pneumonia was significantly increased in patients treated with PD‐1/PD‐L1 inhibitors (ROR, 44.48 [39.05–50.67]; IC, 5.26 [5.04]), CD20 inhibitors (17.45 [13.88–21.94]; 3.95 [3.57]), and HER2 inhibitors (15.95 [11.88–21.40]; 3.76 [3.26]), with the highest risk observed in those receiving PD‐1/PD‐L1 inhibitors. The signal risk of developing eosinophilic showed signals of disproportionate reporting with all drugs except CD20 inhibitors (Figure 3 and Table S3).

FIGURE 3.

Subgroup analysis based on IC with IC_0.25 (A) and ROR with 95% CI (B) values for disproportionality in drug‐associated adverse events across four ILD subtypes. CD20, clusters of differentiation 20; EGFR, epidermal growth factor receptor; HER2, human epidermal growth factor receptor 2; ILD, interstitial lung disease; PD‐1/PD‐L1, programmed cell death protein receptor; VEGF/VEGFR, vascular endothelial growth factor.

3.3. Cumulative Report Analysis

The reported cases of the four ILD subtypes disproportionately reported with the five drugs included in this study have steadily increased. Interstitial pneumonitis adverse events reported for these drugs surged in 2012. Before 2012, adverse events related to the four ILD subtypes were predominantly reported for CD20 inhibitors; however, after 2012, ILD adverse events reported for PD‐1/PD‐L1 inhibitors and VEGF/VEGFR inhibitors increased, leading to a decreased proportion of ILD adverse events reported for CD20 inhibitors (Figure S1).

4. Discussion

4.1. Key Findings

This study is the first to analyse the signal detection between cancer‐targeted mAbs and ILD subtypes using a global ADR database from 1968 to 2024. Various types of ILD adverse events reported with mAbs have steadily increased, with a sharp rise observed between 2010 and 2020. Interstitial pneumonitis and eosinophilic pneumonia were predominantly reported in the Western Pacific region, organising pneumonia was most common in the Eastern Mediterranean region, and pulmonary fibrosis occurred frequently in the Americas and the European region. The number of reports of four ILD subtypes related to five monoclonal antibodies was higher in males than in females across all subtypes. Clinical outcome assessment indicated that interstitial pneumonitis and organising pneumonia showed relatively favourable recovery rates, whereas pulmonary fibrosis and eosinophilic pneumonia had a higher proportion of non‐recovered or fatal cases. Disproportionality analyses indicated that all drugs showed significant reporting signals for interstitial pneumonitis and pulmonary fibrosis. Interstitial pneumonitis showed particularly strong signals in reports with EGFR inhibitors and PD‐1/PD‐L1 inhibitors, whereas pulmonary fibrosis was disproportionately reported with CD20 inhibitors. Disproportionate reporting signals for organising pneumonia were also observed with CD20 inhibitors, HER2 inhibitors, and PD‐1/PD‐L1 inhibitors, with the strongest signals reported for PD‐1/PD‐L1 inhibitors. Eosinophilic pneumonia showed disproportionate reporting signals with all monoclonal antibodies except CD20 inhibitors.

Since 2010, ILD adverse events reported with five cancer‐targeted mAbs have steadily increased. This trend may be the result of increased use of VEGF/VEGFR inhibitors and the introduction of PD‐1/PD‐L1 inhibitors. VEGF/VEGFR inhibitors have been continuously evolving and increasing in usage [20], while PD‐1/PD‐L1 inhibitors, approved in 2014, have been increasingly utilised across a wide range of cancer treatments [21].

A regional analysis of adverse events showed different reporting patterns across different ILD subtypes. First, interstitial pneumonitis and eosinophilic pneumonia were most frequently reported in Western Pacific region. As most countries in the Western Pacific Region are low‐ and middle‐income developing economies, this phenomenon may have resulted from rapid urbanisation and industrialization–related particulate air pollution [22]. Second, the rising prevalence of atopic diseases in East Asia may have contributed to higher reporting of eosinophilic pneumonia [23]. Third, the higher organising pneumonia reporting numbers in the Eastern Mediterranean region could be influenced by the widespread presence of deserts which contain harmful substances that can trigger lung inflammation and may led to increased reporting of organising pneumonia [24]. Previous studies have shown that the southeastern Mediterranean region frequently experiences desert dust storms originating from the Sahara and the Arabian Peninsula. These events significantly increase the concentration of air pollutants and have been found to adversely affect cardiopulmonary health [25]. Finally, pulmonary fibrosis reported in the Americas and the European region, possibly due to these regions serving as central hubs for cancer‐targeted therapy research [26, 27] and the high occurrence of MUC5B gene mutations [28].

The number of reported ILD adverse events was higher in males than in females. This observation may reflect the influence of the higher global smoking prevalence among males (32.6% in 2022) compared with females (6.5%) [29]. Moreover, although still hypothetical, the potential protective effect of oestrogen in attenuating pulmonary fibrosis may also have contributed [30]. Clinically, pulmonary fibrosis and eosinophilic pneumonia had particularly poor prognoses. In progressive cases of pulmonary fibrosis, the irreversible progression of fibrosis makes recovery impossible, and various complications may further worsen the prognosis [31]. Additionally, eosinophilic pneumonia can cause severe lung damage due to drug‐induced eosinophilic infiltration, and its non‐specific symptoms may lead to delayed diagnosis, resulting in a worse prognosis [32, 33]. However, eosinophilic pneumonia is generally known to be more responsive to steroid treatment compared to other forms of interstitial pneumonitis [34]. Therefore, the high fatality rate observed in this study may be attributed to the relatively small number of eosinophilic pneumonia cases, which could have resulted in a proportionally higher fatality rate. Additionally, this finding may reflect the prognosis of the underlying malignancy rather than the adverse event itself. Most eosinophilic pneumonia cases were observed with VEGF/VEGFR inhibitors, commonly prescribed for solid tumours. In contrast, pulmonary fibrosis was more frequently reported with CD20 inhibitors commonly used in hematologic malignancies. These differences in underlying comorbid conditions may have contributed to the observed disparity in fatality.

Across all drugs, disproportionality analyses indicated signals for interstitial pneumonitis and pulmonary fibrosis. Interstitial pneumonitis showed a particularly strong reporting signals with EGFR inhibitors and PD‐1/PD‐L1 inhibitors, while pulmonary fibrosis was more frequently reported with CD20 inhibitors. Although the exact mechanisms have not been fully elucidated, EGFR inhibitors impair EGF‐induced alveolar epithelial cell recovery [35], and studies in animal models have shown that these inhibitors exacerbate lung injury and sustain pulmonary inflammation [36]. Similarly, PD‐1/PD‐L1 inhibitors may induce inflammatory responses by disrupting the regulation of immune effector cells and T cells within the pulmonary interstitium, potentially contributing to interstitial pneumonitis development [37]. CD20 inhibitors may contribute to pulmonary fibrosis development by inducing immune dysregulation, promoting cytokine release, and activating the complement system, thereby leading to lung damage [1, 3].

An increased reporting signal for organising pneumonia was observed with CD20 inhibitors, HER2 inhibitors, and PD‐1/PD‐L1 inhibitors, with the strongest signal detected in those receiving PD‐1/PD‐L1 inhibitors. The immune dysregulation attributed to PD‐1/PD‐L1 inhibitors may induce pulmonary toxicity, which could ultimately lead to the development of organising pneumonia [38]. Lastly, eosinophilic pneumonia exhibited pharmacovigilance signals with all cancer‐targeted mAbs except for CD20 inhibitors. Overall, cancer‐targeted mAbs have been suggested to generate pharmacovigilance signals of pulmonary adverse events, potentially explained by mechanisms such as complement activation and cytokine release, which may help account for the disproportionate reporting of eosinophilic pneumonia [39]. Furthermore, cancer‐targeted monoclonal antibodies have been proposed to contribute to alveolar epithelial injury through direct cytotoxic and immune‐mediated mechanisms. When such injury recurs, it may provoke dysregulated wound‐healing processes, ultimately leading to pulmonary fibrosis characterised by excessive deposition of extracellular matrix [40].

Furthermore, cancer‐targeted mAbs have been suggested to trigger inflammatory responses in pulmonary tissue through direct cytotoxic effects or immune‐mediated mechanisms [41]. When such inflammation is prolonged or severe, the subsequent healing process has been proposed as a possible explanation for signals of scar tissue formation and progression to fibrosis [1]. Accordingly, inflammation and fibrosis may be interpreted as sequential stages that could underlie the reporting patterns observed in this study.

4.2. Comparison of Previous Studies

Due to their therapeutic efficacy, cancer‐targeted mAbs have become increasingly prominent globally. However, as previous studies have documented safety concerns, questions have arisen about their potential risks, including ILD. Consistent with findings from previous studies, the emergence of mAbs has been linked to increased pharmacovigilance signals of ILD [3, 12, 42]. Specifically, reporting signals for ILD have been observed with PD‐1/PD‐L1 inhibitors [43, 44], HER2 inhibitors [45], VEGF‐VEGFR inhibitors [11], and EGFR inhibitors, particularly among males [44, 46]. However, these studies have several limitations, such as poor quality of the included studies, which affected their reliability [7], the lack of distinction between ILD subtypes [3, 12, 42, 43, 45, 46], the inclusion of a relatively small sample size (n = 186 [33], n = 1541 [36] vs. n = 5941 of our study), and the focus on only one or two specific drugs [11, 43, 44, 45, 46]. To address these limitations, our study focused on four distinct ILD subtypes, utilising a large‐scale global pharmacovigilance system and analysing five different types of cancer‐targeted mAbs through disproportionality‐based signal detection analyses.

Furthermore, several studies addressing regional differences in the reporting patterns between mAb types and ILD subtypes align with our analysis. For instance, higher reporting frequencies of interstitial pneumonitis and eosinophilic pneumonia in East Asia [23, 47, 48], organising pneumonia in the Eastern Mediterranean [24], and pulmonary fibrosis in the Americas and Europe [26, 27, 28] has been documented. In addition, numerous studies have noted that ILD was more frequently reported in males in most cases [49, 50, 51], which is consistent with our analysis. However, these studies primarily focused on individual countries, resulting in a limited global perspective. To address this gap, this study aimed to utilise a global drug database for research.

4.3. Clinical Policy Implications

To the best of our knowledge, previous studies have not sufficiently considered ILD subtypes in their analyses. Therefore, there is an urgent need to strengthen international databases to enable more precise signal detection of these subtypes. To address this, it is essential to mandate healthcare professionals worldwide to promptly report adverse reactions when they occur, and there should be an agreed template for nomenclature in such reports to avoid misclassification.

Most of the ILD subtypes reported with the mAbs can exhibit improved outcomes if diagnosed early and prompt discontinuation of the suspected drug. However, the clinical similarities between these conditions and other pulmonary diseases often delay diagnosis and identification of underlying causes, leading to poorer prognoses. Thus, it is essential to develop clinical guidelines that address the disproportionate reporting signals of specific ILD subtypes observed with particular drugs. Additionally, a structured monitoring system should be implemented for which pharmacovigilance signals of ILD have been identified. These measures will help ensure early diagnosis and timely medical intervention, ultimately enhancing patient safety and improving clinical outcomes.

4.4. Limitations

This study is the first to utilise global ADR data to compare and analyse signal detection between cancer‐targeted mAbs and ILD subtypes. However, several limitations should be acknowledged. First, due to the self‐reporting nature of the dataset, this study may underestimate the actual occurrence of adverse events, as it includes only spontaneously reported cases rather than all individuals exposed to monoclonal antibodies [52]. Furthermore, because the data used in this study are based on spontaneous reports rather than systematic surveys, key confounders such as demographic factors were not fully captured. Consequently, we are unable to assess the impact of alcohol consumption and smoking status on the outcomes. Second, this study has the limitation of being a cross‐sectional study. While a cross‐sectional design could indicate reporting patterns between drugs and ILD, it is not sufficient to establish direct causality. Because VigiBase is based on spontaneous reporting, although it cannot be entirely excluded, only a minimal number of cases of ILD exacerbations in patients with pre‐existing disease are expected to have been misclassified as adverse drug reactions. VigiBase applies systematic validation processes to identify and exclude incomplete or implausible reports, which further reduces the likelihood of misclassification [53]. To minimise this bias, we excluded cases coded as secondary ILDs in MedDRA (e.g., respiratory bronchiolitis‐associated ILD, rheumatoid arthritis‐associated ILD) and restricted our dataset to drugs designated as suspected in the drug role field while excluding those recorded exclusively as concomitant, consistent with standard practices in VigiBase analyses [54]. Within this framework, our disproportionality analysis was designed to detect disproportionate reporting signals between cancer‐targeted monoclonal antibodies and ILD subtypes, and the ROR and IC metrics should therefore be interpreted as indicators of signal detection rather than as evidence of incidence or causality [55]. Third, this study has the limitation of being unable to account for factors other than the drugs that may have influenced adverse events. The dataset used in this study did not include health information beyond drug‐related data. As such, factors unrelated to drug exposure may have contributed to the adverse events, potentially influencing the results. Fourth, diagnosing ILD is challenging due to its numerous subtypes, making it difficult to classify drug‐related reports with precision. However, we carefully excluded idiopathic and cryptogenic cases based on well‐established studies to minimise bias in case selection. Additionally, most reported cases were diagnosed by healthcare professionals, ensuring a high level of reliability in our analysis. Nevertheless, it remains unclear whether the diagnoses in this study were made according to standard clinical procedures or through consensus by an ILD multidisciplinary team, and therefore, caution is warranted when interpreting the results. Finally, as ILDs remain poorly defined, ongoing research continues to explore their pathogenesis and treatment, making subclassification challenging. However, to minimise bias from this circumstance, we referred to established studies when selecting MedDRA terms and ATC codes for this analysis [17, 18]. Fifth, the dataset used in this study was derived from individual ICSR coding rather than from a centrally adjudicated review process; consequently, the classification of ILD subtypes may have varied across reporting institutions. Although the majority of reports were submitted by healthcare professionals, the absence of central adjudication raises the possibility of inconsistencies in subtype assignment. Therefore, these limitations should be carefully considered when interpreting the results. Sixth, the dataset used in this study did not include information on patients' underlying cancer diagnoses or other comorbid conditions. These clinical characteristics are important potential confounders, as the type and prognosis of the underlying malignancy may influence fatality outcomes independently of the adverse drug reaction itself. The inability to adjust for these variables in the statistical models limits the capacity to isolate the specific contribution of pharmacovigilance signals related to ILD subtypes to observed fatality. Despite these limitations, by utilising a long‐term and extensive global pharmacovigilance system, we have identified disproportionate reporting signals between cancer‐targeted mAbs and ILD subtypes.

In conclusion, this study is the first to analyse the signal detection of cancer‐targeted mAbs in relation to ILD subtypes using a global pharmacovigilance system, an ADR database from 1968 to 2024. The findings indicated a continuous increase in ILD adverse events, with a significant surge observed since 2010 due to the increased use of VEGF/VEGFR inhibitors and the introduction of PD‐1/PD‐L1 inhibitors. Furthermore, disproportionate reporting signals were identified between cancer‐targeted mAbs and ILD subtypes, with specific drugs exhibiting distinct signal patterns across different ILD subtypes. Although this study did not allow causal interpretation, it underscores the need to strengthen ADR reporting systems, establish clinical guidelines informed by pharmacovigilance signals of ILD subtypes, and develop systematic monitoring protocols. These measures will facilitate early detection and appropriate clinical interventions, ultimately enhancing patient safety.

Author Contributions

Jinyoung Jeong: conceptualization (equal), formal analysis (equal), resources (equal), software (equal), supervision (equal), validation (equal), visualization (equal), writing – original draft (equal), writing – review and editing (equal). Hyunjee Kim: conceptualization (equal), data curation (equal), formal analysis (equal), funding acquisition (equal), investigation (equal), methodology (equal), project administration (equal), resources (equal), software (equal), supervision (equal), validation (equal), visualization (equal), writing – original draft (equal), writing – review and editing (equal). Sooji Lee: conceptualization (equal), data curation (equal), formal analysis (equal), funding acquisition (equal), investigation (equal), methodology (equal), project administration (equal), resources (equal), software (equal), supervision (equal), validation (equal), visualization (equal), writing – original draft (equal), writing – review and editing (equal). Jiyoung Hwang: writing – review and editing (equal). Lee Smith: writing – review and editing (equal). Ho Geol Woo: writing – review and editing (equal). Jaehyeong Cho: conceptualization (equal), data curation (equal), formal analysis (equal), funding acquisition (equal), investigation (equal), methodology (equal), project administration (equal), resources (equal), software (equal), supervision (equal), validation (equal), visualization (equal), writing – original draft (equal), writing – review and editing (equal). Dong Keon Yon: conceptualization (equal), data curation (equal), formal analysis (equal), funding acquisition (equal), investigation (equal), methodology (equal), project administration (equal), resources (equal), software (equal), supervision (equal), validation (equal), visualization (equal), writing – original draft (equal), writing – review and editing (equal).

Ethics Statement

The use and manipulation of the patient data in this study was approved by the Institutional Review Board of Kyung Hee University. Furthermore, as this study is based on population‐level data, ethical approval and informed consent were not required.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: MedDRA terms and classifications for subtypes of ILD adverse events.

Table S2: Analysis of drug‐associated ILD adverse events stratified by age based on IC and IC_0.25 values.

Table S3: Analysis of drug‐associated ILD adverse events across subgroups with ROR (95% CI) and IC (IC_0.25).

Figure S1: (A) Annual reported incidence of ILD adverse events by reporting year and (B) cumulative reports of drug‐associated ILD adverse events by year. CD20, clusters of differentiation 20; EGFR, epidermal growth factor receptor; HER2, human epidermal growth factor receptor 2; ILD, interstitial lung disease; PD‐1/PD‐L1, programmed cell death protein receptor; VEGF/VEGFR, vascular endothelial growth factor.

Jeong J., Kim H., Lee S., et al., “Global Disproportionality Analysis of Adverse Event Reports on Interstitial Lung Diseases for Cancer‐Targeted Therapies, 1968–2024,” Respirology 31, no. 2 (2026): 152–161, 10.1002/resp.70135.

Data Availability Statement

Pharmacovigilance data (VigiBase) are available on reasonable request. Study protocol, statistical code: available from D.K.Y. (email: yonkkang@gmail.com). Data set: available from the Uppsala Monitoring Centre or World Health Organization through a data use agreement.